Ebffiledoc 3819

Machine Learning and Intelligent
Communications Third International

Conference MLICOM 2018 Hangzhou
China July 6 8 2018 Proceedings Limin
Meng
Visit to download the full and correct content document:
https://textbookfull.com/product/machine-learning-and-intelligent-communications-thir
d-international-conference-mlicom-2018-hangzhou-china-july-6-8-2018-proceedings-li
min-meng/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Machine Learning and Intelligent Communications Second

International Conference MLICOM 2017 Weihai China
August 5 6 2017 Proceedings Part II 1st Edition Xuemai
Gu
https://textbookfull.com/product/machine-learning-and-
intelligent-communications-second-international-conference-
mlicom-2017-weihai-china-august-5-6-2017-proceedings-part-ii-1st-
edition-xuemai-gu/
Service Oriented Computing 16th International

Conference ICSOC 2018 Hangzhou China November 12 15
2018 Proceedings Claus Pahl
https://textbookfull.com/product/service-oriented-computing-16th-
international-conference-icsoc-2018-hangzhou-china-
november-12-15-2018-proceedings-claus-pahl/
Intelligent Information Processing X 11th IFIP TC 12

International Conference IIP 2020 Hangzhou China July 3
6 2020 Proceedings Zhongzhi Shi
https://textbookfull.com/product/intelligent-information-
processing-x-11th-ifip-tc-12-international-conference-
iip-2020-hangzhou-china-july-3-6-2020-proceedings-zhongzhi-shi/
Machine Learning and Intelligent Communications First

International Conference MLICOM 2016 Shanghai China
August 27 28 2016 Revised Selected Papers 1st Edition
Huang Xin-Lin (Eds.)
https://textbookfull.com/product/machine-learning-and-
intelligent-communications-first-international-conference-
mlicom-2016-shanghai-china-august-27-28-2016-revised-selected-
Cooperative Design Visualization and Engineering 15th
International Conference CDVE 2018 Hangzhou China
October 21 24 2018 Proceedings Yuhua Luo
https://textbookfull.com/product/cooperative-design-
visualization-and-engineering-15th-international-conference-
cdve-2018-hangzhou-china-october-21-24-2018-proceedings-yuhua-
luo/
Web Information Systems and Applications: 15th

International Conference, WISA 2018, Taiyuan, China,
September 14–15, 2018, Proceedings Xiaofeng Meng
https://textbookfull.com/product/web-information-systems-and-
applications-15th-international-conference-wisa-2018-taiyuan-
china-september-14-15-2018-proceedings-xiaofeng-meng/
Intelligent Systems Design and Applications: 18th

International Conference on Intelligent Systems Design
and Applications (ISDA 2018) held in Vellore, India,
December 6-8, 2018, Volume 1 Ajith Abraham
https://textbookfull.com/product/intelligent-systems-design-and-
applications-18th-international-conference-on-intelligent-
systems-design-and-applications-isda-2018-held-in-vellore-india-
december-6-8-2018-volume-1-ajith-abraham/
Intelligent Systems Design and Applications: 18th

International Conference on Intelligent Systems Design
and Applications (ISDA 2018) held in Vellore, India,
December 6-8, 2018, Volume 2 Ajith Abraham
https://textbookfull.com/product/intelligent-systems-design-and-
applications-18th-international-conference-on-intelligent-
systems-design-and-applications-isda-2018-held-in-vellore-india-
december-6-8-2018-volume-2-ajith-abraham/
ICCCE 2018: Proceedings of the International Conference

on Communications and Cyber Physical Engineering 2018
Amit Kumar
https://textbookfull.com/product/iccce-2018-proceedings-of-the-
international-conference-on-communications-and-cyber-physical-
engineering-2018-amit-kumar/
Limin Meng
Yan Zhang (Eds.)
251
Machine Learning
and Intelligent
Communications
Third International Conference, MLICOM 2018
Hangzhou, China, July 6–8, 2018
Proceedings
123
Lecture Notes of the Institute
for Computer Sciences, Social Informatics
and Telecommunications Engineering 251
Editorial Board
Ozgur Akan
Middle East Technical University, Ankara, Turkey
Paolo Bellavista
University of Bologna, Bologna, Italy
Jiannong Cao
Hong Kong Polytechnic University, Hong Kong, Hong Kong
Geoffrey Coulson
Lancaster University, Lancaster, UK
Falko Dressler
University of Erlangen, Erlangen, Germany
Domenico Ferrari
Università Cattolica Piacenza, Piacenza, Italy
Mario Gerla
UCLA, Los Angeles, USA
Hisashi Kobayashi
Princeton University, Princeton, USA
Sergio Palazzo
University of Catania, Catania, Italy
Sartaj Sahni
University of Florida, Florida, USA
Xuemin Sherman Shen
University of Waterloo, Waterloo, Canada
Mircea Stan
University of Virginia, Charlottesville, USA
Jia Xiaohua
City University of Hong Kong, Kowloon, Hong Kong
Albert Y. Zomaya
University of Sydney, Sydney, Australia
More information about this series at http://www.springer.com/series/8197
Limin Meng Yan Zhang (Eds.)
•
Machine Learning
and Intelligent
Communications
Third International Conference, MLICOM 2018
Hangzhou, China, July 6–8, 2018
Proceedings
123
Editors
Limin Meng Yan Zhang
Zhejiang University of Technology Department of Networks
Hangzhou Shi Simula Research Laboratory
China Oslo
Norway
ISSN 1867-8211 ISSN 1867-822X (electronic)

Lecture Notes of the Institute for Computer Sciences, Social Informatics
and Telecommunications Engineering
ISBN 978-3-030-00556-6 ISBN 978-3-030-00557-3 (eBook)
https://doi.org/10.1007/978-3-030-00557-3
Library of Congress Control Number: 2018954076
© ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2018
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
We are delighted to introduce the proceedings of the 2018 European Alliance for
Innovation (EAI) International Conference on Machine Learning and Intelligent
Communications (MLICOM 2018). This conference has brought together researchers,
developers, and practitioners around the world who are leveraging and developing
potential intelligent solutions for future mobile communications and networks. We
hope the integrating of machine learning algorithms into communication systems will
improve the QoS and make the systems smart, intelligent, and efficient.
The technical program of MLICOM 2018 consisted of 66 full papers in oral pre-
sentation sessions at the main conference tracks. The conference tracks were: Track 1,
Machine learning; Track 2, Intelligent positioning and navigation; Track 3, Intelligent
Multimedia Processing and Security; Track 4, Wireless Mobile Network and Security;
Track 5, Cognitive Radio and Intelligent Networking; Track 6, Intelligent Internet of
Things; Track 7, Intelligent Satellite Communications and Networking; Track 8,
Intelligent Remote Sensing, Visual Computing and Three-Dimensional Modeling;
Track 9, Green Communication and Intelligent Networking; Track 10, Intelligent
Ad-hoc and Sensor Networks; Track 11, Intelligent Resource Allocation in Wireless
and Cloud Networks; Track 12, Intelligent Cooperative Communications and Net-
working; Track 13, Intelligent Computing and Caching; Track 14, Intelligent Signal
Processing in Wireless and Optical Communications; Track 15, Intelligent Radar
Signal Processing. Aside from the high-quality technical paper presentations, the
technical program also featured four keynote speeches. The four keynote speeches were
Prof. Xuemin (Sherman) Shen from the University of Waterloo, Canada, Prof. Nei
Kato from the Tohoku University, Japan, Prof. Qinyu Zhang from the Harbin Institute
of Technology, China, and Prof. Liping Qian from the Zhejiang University of Tech-
nology, China.
Coordination with the steering chairs, Prof. Imrich Chlamtac, Prof. Xin Liu, and
Prof. Xin-Lin Huang, was essential for the success of the conference. We sincerely
appreciate their constant support and guidance. It was also a great pleasure to work
with such an excellent Organizing Committee team who worked hard to organize and
support the conference and, in particular, the Technical Program Committee, led by our
TPC co-chairs, Prof. Yan Zhang, Prof. Bo Ji, Prof. Shuai Han, and Prof. Liping Qian
who completed the peer-review process of technical papers and created a high-quality
technical program. We are also grateful to the conference managers, Katarina Antalova,
Radka Pincakova, and Jana Štefaňáková, for their support and to all the authors who
submitted their papers to the MLICOM 2018 conference.
We strongly believe that MLICOM conference provides a good forum for all
researcher, developers, and practitioners to discuss all science and technology aspects
VI Preface
that are relevant to machine learning and intelligent communications. We also expect
that the future MLICOM conferences will be as successful and stimulating, as indicated
by the contributions presented in this volume.
August 2018 Limin Meng

Yan Zhang
Organization
Steering Committee
Steering Committee Chair
Imrich Chlamtac University of Trento, Italy, and EAI
Steering Committee Co-chair

Xin Liu Dalian University of Technology, China
Steering Committee Member

Xin-Lin Huang Tongji University, China
Organizing Committee
General Chairs
Limin Meng Zhejiang University of Technology, China
Yan Zhang University of Oslo, Norway
General Co-chairs
Su Hu University of Electronic Science and Technology of China,
China
Yuan Wu Zhejiang University of Technology, China
Weidang Lu Zhejiang University of Technology, China
Technical Program Committee Chairs

Yan Zhang University of Oslo, Norway
Bo Ji Temple University, USA
Shuai Han Harbin Institute of Technology, China
Liping Qian Zhejiang University of Technology, China
Web Chair
Liang Huang Zhejiang University of Technology, China
Publicity and Social Media Chair

Qing Pan Zhejiang University of Technology, China
Sponsorship and Exhibits Chair

Hong Peng Zhejiang University of Technology, China
VIII Organization
Publications Chair
Yu Zhang Zhejiang University of Technology, China
Local Chairs
Conference Manager
Katarina Antalova EAI - European Alliance for Innovation
Technical Program Committee

Machine Learning
Xinlin Huang Tongji University, China
Huasen Wu Twitter, USA
Intelligent Positioning and Navigation

Mu Zhou Chongqing University of Posts and Telecommunications,
China
Zhian Deng Harbin Engineering University, China
Zhiguo Shi Zhejiang University, China
Intelligent Multimedia Processing and Security

Fen Hou University of Macau, Macau, SAR China
Hangguan Shan Zhejiang University, China
Wireless Mobile Network and Security

Kuan Zhang University of Nebraska-Lincoln, USA
Shijun Lin Xiamen University, China
Xiaojie Fang Harbin Institute of Technology, China
Cognitive Radio and Intelligent Networking

Xin Liu Dalian University of Technology, China
Huiming Wang Xian Jiaotong University, China
Intelligent Internet of Things

Kaikai Chi Zhejiang University of Technology, China
Yongliang Sun Nanjing Tech University, China
Ruilong Deng University of Alberta, Canada
Organization IX
Intelligent Satellite Communications and Networking

Kanglian Zhao Nanjing University, China
Haibo Zhou Nanjing University, China
Shaohua Wu Harbin Institute of Technology, China
Intelligent Remote Sensing, Visual Computing and Three-Dimensional Modeling

Shifang Xu Shanghai Institute of Technology, China
Guodao Sun Zhejiang University of Technology, China
Green Communication and Intelligent Networking

Yu Zhang Zhejiang University of Technology, China
Nan Zhao Dalian University of Technology, China
Intelligent Ad-hoc and Sensor Networks

Jian Qiu Hangzhou Dianzi University, China
Zhenyu Na Dalian Maritime University, China
Danyang Qin Heirongjiang University, China
Intelligent Resource Allocation in Wireless and Cloud Networks

Caihong Kai Hefei University of Technology, China
Nan Cheng University of Waterloo, Canada
Suzhi Bi Shenzhen University, China
Intelligent Cooperative Communications and Networking

Deli Qiao East China Normal University, China
Xuewen Liao Xi’an Jiaotong University, China
Shan Zhang University of Waterloo, Canada
Min Jia Harbin Institute of Technology, China
Intelligent Computing and Caching

Ju Ren Central South University, China
Juan Liu Ningbo University, China
Wei Wang Zhejiang University, China
Zheng Chang University of Jyvaskyla, Finland
Intelligent Signal Processing in Wireless and Optical Communications

Qingjiang Shi Nanjing University of Aeronautics and Astronautics, China
Tong Ye Shanghai Jiaotong Univeristy, China
Shaohua Hong Xiamen University, China
Jing Xu Zhejiang University, China
X Organization
Intelligent Radar Signal Processing

Yujie Gu Temple University, USA
Weijie Xia Nanjing University of Aeronautics and Astronautics, China
Xiaolong Chen Naval Aviation University, China
Contents
Machine Learning
Gun Identification Using Tensorflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Mitchell Singleton, Benjamin Taylor, Jacob Taylor, and Qingzhong Liu
A Research of Network Applications Classification Based

on Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Hong Shao, Liujun Tang, Ligang Dong, Long Chen, Xian Jiang,
and Weiming Wang
Real-Time Drone Detection Using Deep Learning Approach . . . . . . . . . . . . 22

Manjia Wu, Weige Xie, Xiufang Shi, Panyu Shao, and Zhiguo Shi
Deep Reinforcement Learning-Based Task Offloading and Resource

Allocation for Mobile Edge Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Liang Huang, Xu Feng, Liping Qian, and Yuan Wu
RFID Data-Driven Vehicle Speed Prediction Using Adaptive

Kalman Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Anqi Feng, Liping Qian, and Yupin Huang
Speed Prediction of High Speed Mobile Vehicle Based on Extended

Kalman Filter in RFID System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Yupin Huang, Liping Qian, and Anqi Feng
Mobility Prediction Based on POI-Clustered Data . . . . . . . . . . . . . . . . . . . . 60

Haoyuan Chen, Yali Fan, Jing Jiang, and Xiang Chen
A Research on the Identification of Internet User Based on Deep

Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Hong Shao, Liujun Tang, Ligang Dong, Long Chen, Xian Jiang,
and Weiming Wang
Intelligent Positioning and Navigation
An Improved Constrained Least Squares Localization Algorithm in NLOS

Propagating Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Yejia Yin, Jingyu Hua, Fangni Chen, Weidang Lu, Dongming Wang,
and Jiamin Li
XII Contents
Intelligent Multimedia Processing and Security
Fast Inter Prediction Mode Decision Algorithm Based on Data Mining . . . . . 95

Tengrui Shi, Xiaobo Guo, Daihui Mo, and Jian Wang
Wireless Mobile Network and Security
Smartphone Application Identification by Convolutional Neural Network . . . . 105

Shuang Zhao and Shuhui Chen
Joint D2D Cooperative Relaying and Friendly Jamming Selection

for Physical Layer Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Yijie Luo, Yang Yang, Yanlei Duan, and Zhengju Yang
Variable Tap-Length Blind Equalization for Underwater Acoustic

Communication. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Zhiyong Liu, Yinyin Wang, and Fan Bai
Random Sequence Generation Algorithm for Multi-chaotic Systems . . . . . . . 136

Xiaodi Chen and Hong Wu
VulAware: Towards Massive-Scale Vulnerability Detection in Cyberspace . . . 144

Zhiqiang Wang, Pingchuan Ma, Ruming Wang, Shichun Gao,
Xuying Zhao, and Tao Yang
Cognitive Radio and Intelligent Networking
Optimization in Cognitive Radio Networks with SWIPT-Based DF Relay . . . 153

Jie Zhang, Weidang Lu, Hong Peng, Zhijiang Xu, and Xin Liu
Real-Valued Orthogonal Sequences for Ultra-low Overhead Channel

Estimation in MIMO-FBMC Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Su Hu, Jing Zhang, Wanbin Tang, Zilong Liu, Pei Xiao, and Yuan Gao
A Novel Spectrum Allocation Scheme in Femtocell Networks Using

Improved Graph Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Feng Li, Lili Yang, Jiangxin Zhang, and Li Wang
Distributed Channel Allocation for Wireless Mesh Networks

Based on Hypergraph Interference Model. . . . . . . . . . . . . . . . . . . . . . . . . . 181
Chen Pan, Yunpeng Cheng, Ducheng Wu, Lei Zhao, and Yuli Zhang
Physical Violence Detection with Movement Sensors . . . . . . . . . . . . . . . . . 190

Liang Ye, Le Wang, Peng Wang, Hany Ferdinando, Tapio Seppänen,
and Esko Alasaarela
Contents XIII
Joint Resource Allocation for Wireless Energy Harvesting Based

on DF Relaying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Tian Nan, Weidang Lu, and Zhijiang Xu
Optimal Power Splitting of Cognitive Radio Networks with

SWIPT-Enabled Relay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Yuanrong Lin, Weidang Lu, Hong Peng, and Jingyu Hua
Deflection Angle Detection of the Rotor and Signal Processing

for a Novel Rotational Gyroscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Dianzhong Chen and Zhongzhao Zhang
Intelligent Internet of Things
Reusing Wireless Power Transfer for Backscatter-Assisted Cooperation

in WPCN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Wanran Xu, Suzhi Bi, Xiaohui Lin, and Juan Wang
Intelligent Satellite Communications and Networking
Channel Estimation in Next Generation LEO Satellite Communicastion

Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Zheng Pan, Zhenyu Na, Xin Liu, and Weidang Lu
A Novel Approach of Semi-blind Frequency Selection for HF Regional

Emergency Maneuver Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Dai-hui Mo, Guo-jun Li, Xiao-fei Xu, Lu Tan, and Ya-kun Xing
Green Communication and Intelligent Networking
A Network Coding Optimization Algorithm for Reducing

Encoding Nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Limin Meng and Yangtianxiu Hu
Backscatter-Assisted Relaying in Wireless Powered Communications

Network. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Yuan Zheng, Suzhi Bi, and Xiaohui Lin
Energy-Efficient Power Allocation Scheme Based on Discrete-Rate

Adaptive Modulation in Distributed Antenna System . . . . . . . . . . . . . . . . . . 284
Xi Wang, Xiangbin Yu, Tao Teng, and Guangying Wang
A Method of Balanced Sleep Scheduling in Renewable Wireless

Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
Maohan Song, Weidang Lu, Hong Peng, Zhijiang Xu, and Jingyu Hua
XIV Contents
Intelligent Ad-hoc and Sensor Networks
Pedestrian Walking Model for Floor Plan Building Based

on Crowdsourcing PDR Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Guangda Yang, Yongliang Zhang, Lin Ma, and Leqi Tang
Weight Matrix Analysis Algorithm for WLAN Indoor Positioning System . . . 317
Lin Ma, Jian Li, and Yubin Xu
Manifold Alignment-Based Radio Map Construction in Indoor

Localization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
Ping Ji, Danyang Qin, Pan Feng, and Yan Zhang
Research on the Contextual Information in Scene Classification . . . . . . . . . . 338

Pan Feng, Danyang Qin, Ping Ji, and Jingya Ma
Intelligent Resource Allocation in Wireless and Cloud Networks
OFDM Based SWIPT in a Two-Way Relaying Network . . . . . . . . . . . . . . . 349

Weilin Zhao, Weidang Lu, Hong Peng, Zhijiang Xu, and Jingyu Hua
Dynamic Network Access for Multi-UAV Networks: A Cloud-Assisted

Learning Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
Xiaodu Liu, Yitao Xu, Yanlei Duan, Dianxiong Liu, and Zhiyong Du
Probability-Based Routing Symmetry Metrics . . . . . . . . . . . . . . . . . . . . . . . 370

Qin Wang, Fang Dong, Xin-Li Yang, and Rui Yin
Dynamic Opportunistic Spectrum Access with Channel Bonding

in Mesh Networks: A Game-Theoretic Approach . . . . . . . . . . . . . . . . . . . . 381
Chen Pan, Yunpeng Cheng, Zhengju Yang, and Yuli Zhang
Secrecy Performance Analysis of SWIPT System Based on OFDM

Assisted Interference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
Mei Qin, Weidang Lu, Hong Peng, and Zhijiang Xu
Contract Theory Based on Wireless Energy Harvesting with Transmission

Performance Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
Chen Liu, Hong Peng, Weidang Lu, Zhijiang Xu, and Jingyu Hua
Intelligent Signal Processing in Wireless and Optical Communications
Probabilistic Sorting Memory Constrained Tree Search Algorithm

for MIMO System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 411
Xiaoping Jin, Zheng Guo, Ning Jin, and Zhengquan Li
Contents XV
Intelligent Cooperative Communications and Networking
Uplink Transmission Scheme Based on Rateless Coding in Cloud-RAN . . . . 429

Lingjie Xie, Yu Zhang, Yefan Zhang, Jingyu Hua, and Limin Meng
Rateless Coded Downlink Transmission in Cloud Radio Access Network . . . 438

Hong Peng, Yefan Zhang, Yu Zhang, Lingjie Xie, and Limin Meng
The Second Round
WiFi/PDR Integrated System for 3D Indoor Localization . . . . . . . . . . . . . . . 451

Mu Zhou, Maxim Dolgov, Yiyao Liu, and Yanmeng Wang
Highly-Available Localization Techniques in Indoor Wi-Fi Environment:

A Comprehensive Survey. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
Mu Zhou, Oyungerel Bulgantamir, and Yanmeng Wang
Detecting Phishing Websites with Random Forest . . . . . . . . . . . . . . . . . . . . 470

Shinelle Hutchinson, Zhaohe Zhang, and Qingzhong Liu
Detecting Digital Photo Tampering with Deep Learning. . . . . . . . . . . . . . . . 480

Mitchell B. Singleton and Qingzhong Liu
Joint Time and Power Allocations for Uplink Nonorthogonal Multiple

Access Networks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490
Yuan Wu, Cheng Zhang, Kejie Ni, Jiajun Shi, Liping Qian,
Liang Huang, Weidang Lu, and Limin Meng
Performance Evaluation of Ad-hoc Routing Protocols in Hybrid

MANET-Satellite Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
Xiaoye Xie, Jian Wang, Xiaobo Guo, and Xiaogang Wu
An Improved Quadratic Programming LLOP Algorithm for Wireless

Localization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
Guangzhe Liu, Jingyu Hua, Feng Li, Weidang Lu, and Zhijiang Xu
Optimize Bundle Size in Satellite DTN Links with Markov Method . . . . . . . 519
Peiyuan Si, Weidang Lu, Zhijiang Xu, and Jingyu Hua
An Efficient Method for Estimating Time Series Motif Length Using

Sequitur Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 531
Nguyen Ngoc Phien
Optimal User Grouping and Resource Allocation for Downlink

Non-Orthogonal Multiple Access Systems . . . . . . . . . . . . . . . . . . . . . . . . . 539
Xiaoding Wang, Kejie Ni, Xiangxu Chen, Yuan Wu, Liping Qian,
and Liang Huang
XVI Contents
Technologies and Applications of Narrowband Internet of Things . . . . . . . . . 549

Jia Chen, Jiajun Shi, Xiangxu Chen, Yuan Wu, Liping Qian,
and Liang Huang
A SAR Image Fast Stitching Algorithm Based on Machine Learning. . . . . . . 559

Hongyuan Yao, Haipeng Wang, and Xueyuan Lin
B2DASH: Bandwidth and Buffer-Based Dynamic Adaptive Streaming

over HTTP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
Peihan Du, Jian Wang, Xin Wang, and Zufeng Xu
ZigBee-Based Device-Free Wireless Localization in Internet of Things . . . . . 575

Yongliang Sun, Xiaocheng Wang, Xuzhao Zhang, and Xinggan Zhang
Designing of Environmental Information Acquisition and Reconstruction

System Based on Compressed Sensing. . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
Qiuming Zhao, Bo Li, Hongjuan Yang, Gongliang Liu, and Ruofei Ma
The Design of Inter-satellite Link Topology Based on Time-Slot Allocation

Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594
Siqi Li, Bingbing Zhang, Wenjing Kang, Bo Li, Ruofei Ma,
and Gongliang Liu
MOPSO Optimized Radar CBMeMBer Forward-Backward Smoothing

Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 604
Jiazheng Pei, Yong Huang, Yunlong Dong, and Xiaolong Chen
Study of Radar Target Range Profile Recognition Algorithm

Based on Optimized Neural Network. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 617
Xiaokang Guo, Tao Jian, Yunlong Dong, and Xiaolong Chen
Resource Allocation Based Simultaneous Wireless Information and Power

Transfer for Multiuser OFDM Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
Shanzhen Fang, Weidang Lu, Hong Peng, Zhijiang Xu, and Jingyu Hua
Improved RSA Localization Based on the Lagrange Multiplier

Optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
Jiafei Fu, Jingyu Hua, Zhijiang Xu, Weidang Lu, and Jiamin Li
Particle Swarm Optimization Based Location Recommendation for D2D

Communication Underlying LTE Cellular Networks . . . . . . . . . . . . . . . . . . 641
Chiapin Wang, Ming-Hsun Wu, and Te-Sheng Tsai
A Dual SIS Epidemic Model for Virus Spread Analysis in Cluster-Based

Wireless Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652
Shensheng Tang and Chenghua Tang
Contents XVII
Communication-Efficient Decentralized Cooperative Data Analytics

in Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663
Liang Zhao, Zhihua Li, and Shujie Guo
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673

Machine Learning
Gun Identification Using Tensorflow
Mitchell Singleton, Benjamin Taylor, Jacob Taylor,

and Qingzhong Liu(&)
Department of Computer Science, Sam Houston State University,

Huntsville, TX 77341, USA
{mitchellsingleton,benjamin.taylor,
jacobtaylor,liu}@shsu.edu
Abstract. Automatic video surveillance can assist security personnel in the

identification of threats. Generally, security personnel are monitoring multiple
monitors and a system that would send an alert or warning could give the
personnel extra time to scrutinize if a person is carrying a firearm. In this paper,
we utilize Google’s Tensorflow API to create a digital framework that will
identify handguns in real time video. By utilizing the MobileNetV1 Neural
Network algorithm, our system is trained to identify handguns in various ori-
entations, shapes, and sizes, then the intelligent gun identification system will
automatically interpret if the subject is carrying a gun or other objects. Our
experiments show the efficiency of implemented intelligent gun identification
system.
Keywords: Tensorflow Gun detection Video surveillance
1 Introduction and Background
Automatic video surveillance is utilized by many people and organizations in the video
surveillance category of their physical security protocol. CCTV and webcams are used
in many products, from specialized doorbells with cameras and microphones that allow
for two-way communication and recording when an event is triggered, or Closed-
Circuit Television’s (CCTV) which are used to aid security personnel in preventing or
helping mitigate an incident. These systems use video in some form of method to
record what the camera sees. While it may be easy for a homeowner to review a single
system that has 1 or 2 cameras, the task becomes an arduous and tedious even when
combing through multiple video streams if they implement a larger array of cameras.
Depending on an organization’s footprint and their implementation of their physical
security plan, security personnel may be monitoring 100’s of video streams [1, 2].
These streams should be prioritized in order of importance to determine the number of
personnel needed and how many streams can be observed per personnel.
Recently, there has been an increasing number of incidents involving handguns in
various situations. To detect handguns, in [3], several models were utilized to detect
pistols. These models used the VGG-16 based classifier and then classified the
detection of the gun with either the sliding window or the region proposal approach.
The authors went with the region proposal approach due to its faster detection speed –
© ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2018
L. Meng and Y. Zhang (Eds.): MLICOM 2018, LNICST 251, pp. 3–12, 2018.
https://doi.org/10.1007/978-3-030-00557-3_1
4 M. Singleton et al.
but also tested against the sliding window approach. To achieve a real-time outdoor
concealed-object detection with passive millimeter wave imaging [4], the authors used
a passive millimeter wave (MMW) imaging system ran at or around 94 GHz with 1 Hz
frame rate. To attain automatic recognition, they used both global (aligning the body
and the background together) and local (inside of the body is processed) segmentation
levels to find the concealed objects. To automatically detect firearms and knives in a
CCTV image, the authors in [1] used an approach utilizing MPEG-7 visual descriptions
and a principal component analysis (PCA). They took each frame and removed
everything except the foreground of the image which included the erosion and dilation.
To detect weapons in surveillance camera images [5], the authors improved detecting
weapons in CCTV via the Histogram of Oriented Gradients (HOG) method for clas-
sification. In [6], a framework was developed to identify general objects in video using
still image processes along with object tracking through the 4th dimension of time. In
[7], a real-time detection, tracking and classification of natural video was examined.
The method uses background subtraction and maximally stable extremal region
(MSER) detection. However the method didn’t successfully reach real-time. Recently a
method to detect visual gun by using Harris interest point was proposed [2], the author
uses a visual gun detection based on FREAK descriptor to recognize guns. The system
processes an image, performs color segmentation, performs a boundary extraction and
compares against a similarity higher than 50% and then executes an alarm. This
algorithm was accurate 84.26% of the time and shows promise.
In this paper, we focus on detecting guns vs no guns. We propose a detection
method that is built on a lean convolutional neural network optimized for speed. This
paper is organized into four additional sections. Section 2 explains the definitions and
tools where we will explain the CNNs and tools used to train our algorithm. Section 3
describes the methodology and the effectiveness of our algorithm, followed by our
conclusions in Sect. 4.
2 Definitions and Tools
Tensorflow (https://www.tensorflow.org/) is a platform that is based on dataflow graphs

and is useful in training with deep neural networks [8]. Tensorflow was created by
Google and it is used in their native applications and in dozens of machine learning
projects such as: facial recognition, image processing, speech recognition, and
extracting information such as license plate numbers. Tensorboard, a web application,
is a visualization for understanding Tensorflow trainings and monitoring the training
progress in a web browser [9]. It showcases visualizations such as: graphs, scalars, and
images. Tensorboard shows cross entropy and accuracy depicted into a graph for
convenient visualization of training results. Tensorboard can be started from an Ana-
conda prompt and uses port 6006 on the localhost for its process.
OpenCV (https://opencv.org/) is an open source computer vision library that con-
tains image processing functions and over 2,500 algorithms used for things like facial
recognition [10]. OpenCV can accelerate CUDA and OpenCL GPUs. OpenCV sup-
ports deep learning platforms like Tensorflow. OpenCV is built using a layering
Gun Identification Using Tensorflow 5
process. You have the OS, then the languages and algorithms, the core of OpenCV, and
at the bottom the hardware acceleration layer (HAL) [11].
MobileNet is a collection of vision models for Tensorflow that are mobile-oriented
and used to enhance visual recognition [12]. It is essentially a convolutional neural
network (CNN). MobileNet, unlike a typical CNN, separates the convolution into a
3 3 depth wise convolution and a 1 1 pointwise convolution. The reason Mobi-
leNet is used more than a traditional CNN is the computation time is much quicker with
their architecture, however the accuracy can be slightly lower. Depending on the
architecture, such as MobileNet_0.50.224 or 0.50.192, it can change the width mul-
tiplier and the image resolution. The image resolution is the 192 or 224 and the width
multiplier is the .50. This will help the object detection computation time of the image
processing the smaller the image resolution. The width multiplier can be ranged from
0.25, 0.50, 0.75, and 1.0. The image resolution can be 128, 160, 192, or 224.
ImageNet (http://www.image-net.org/) is a very large dataset of images (over 14
million) with about 22,000 categories and are collected by human labelers [13]. Ima-
geNet compiles hundreds of thousands of thumbnails and URLs of images just like
Google, a search engine, does. ImageNet is organized via synsets which is a concept of
WordNet. ImageNet was created to be utilized by researchers as a primary resource for
images.
Anaconda (https://www.anaconda.com/) is used for large quantities of data pro-
cessing, predictive analytics, and scientific computing [14]. It has over 1000 packages
in its repository that can be installed, 150 of which come pre-installed with Anaconda.
It is an open-source distribution of Python used to ease and simplify package admin-
istration and distribution. The packages are managed by a package management system
within anaconda called conda.
Nvidia Compute Unified Device Architecture (CUDA) is utilized by many as their
primary method to process images because it essentially groups the cores of GPUs into
a vector and can then be programmed to decrease processing time on the large sets of
data versus CPUs that do not run on a parallel throughput architecture [15].
Shi-Tomasi corner point detection is an enhancement to the Harris corner detector.
The way Harris corner detection works is that the pixel is calculated and if above a
precise value then it’s marked as a corner. To score it, two eigen values are used to
check if it’s a corner during detection and then inputted into a function which
manipulates them. Shi-Tomasi decided to enhance this method and only rely on the two
eigen values, not the entire Harris formula. This is because their model relies solely on
the tracker’s accuracy [16].
3 Methodologies
The dataset pictures were gathered from the Sun database [17], Internet Movie Firearm
Database [18], Pixabay [19], ImageNet [20] and reduced in number to 3363 gun images
and 11834 non-gun images. The dataset was trained several times adding or removing
pictures from the datasets to improve accuracy.
The method utilized in this paper was to retrain multiple object detection models
(inception_v3, MobileNet_0.50_192, MobileNet_0.50_224, MobileNet_1.0_224) for a
binary classification of an object into either gun or no gun using a training dataset
composed of firearms and non-firearm pictures. See Fig. 1 for the MobileNet archi-
tecture training flow chart.
Fig. 1. MobileNet 1.0 224 training flowchart
The use of Tensorboard allowed watching the progress of the retraining and in
Figs. 2 and 3 the visualization of the progress of the retraining can be seen from step 60
to step 4000 respectively.
Once the retraining was complete, Shi Tomasi key point detection was used to
identify the parts of a picture that should be looked at by the binary classification. The
assumption being that the key points would be less than the number of sliding windows
across the whole picture.
The key point locations are then each used to crop out a section of the image and
the retrained model is called to determine if the cropped section contains a gun or not.
If a gun is found, then the cropped section has a rectangle drawn on the picture in the
same location as the cropping. In the next section, we will review the results of our
current models and some examples of identified pictures in the gun category and
correctly identified pictures in the non-gun category.
Fig. 2. Starting out the retraining
Fig. 3. Finished with the retraining
4 Results
The gun detection was trained and tested on the following datasets described in Sect. 3.
The images detected in a fast method and we would like to expand upon static images
to real time video as our algorithm is improved. MobileNet was trained in various step
counts and tested for accuracy. In one training iteration, we trained 3000 steps as seen
in Table 1 and resulting in 97.89% accuracy. In another iteration, we trained the
MobileNet 1.0 224 model 5000 steps and the accuracy increased to 100%. Our most
accurate training algorithm was trained on 795 guns and 3944 non-guns. Testing using
the results from this training resulted in the least false-positives compared to our dataset
that included more pictures in the gun dataset. Results from this iteration of testing
gave a 96% accuracy when testing against non-gun photos with the guns filtered at
89%. When testing against photos with guns, we have an accuracy of 86.67% out of 30
photos. These calculations can be referenced in Table 1 (Fig. 4).
Table 1. Model detection results

Pictures Correct Incorrect Correct
Non-gun 25 24 1 96%
Gun 30 26 4 86.67%
Fig. 4. MobileNet 1.0 224 retraining results
Our theory is accuracy will go up if we can preprocess the images with by sub-
tracting the background, but this will have to be completed in another project. Some
photos were missed because there is a limit of 25 corners maximum per image.
Increasing this number will increase the amount of time the image must be processed,
but we believe this will be minimized by removing or subtracting the background.
Increasing the quality of our gun data set should improve the accuracy of the gun
detection rate and will allow us to increase from 89% sensitivity without increasing
false alarms, which can lead to alarm fatigue. We have included examples of correctly
identified guns in Figs. 5, 6, and 7. The detection algorithm perform a gray transform.
In Fig. 5, there were three positive hits. The background was uniform and the detection
could identify the gun at least 3 times. It also correctly did not detect a gun on the
subject’s face, hair, earing, or shoulder. In Fig. 6, the subject had multiple hits on
Fig. 5. Successful detection of a gun

various parts of the gun. In future iterations of the code, we will want to decrease the
output of overlapping gun identifiers. If it finds multiple positive hits within x pixels,
then either combine or ignore x screen draws.
In Fig. 7, you will see multiple identifiers on the base of the gun, on the barrel of
the gun, and on a pistol like device in the bottom left corner. The background is the
most difficult of the three different gun detection examples as the bricks do have similar
hard angles that can be found in a gun. None of the bricks were identified as a gun so
we are happy to report the background training with multiple forms of background
photos in non-gun dataset has improved upon our earlier training and results. Again, we
believe minimizing the number of identified boxes will clean up the display and will
aid in the time of displaying the results in real time when this algorithm is improved
and created for a real-time video system.
We have also included correctly identified photos with no guns in Figs. 8 and 9. In
Fig. 8, there are multiple hard edges in the grill and around the tires that have a gun
shape. The algorithm was smart enough to not incorrectly give a false positive in all the
dots that were run against. Other thing to note in the picture is the background was
uniform with only having shadows at the bottom of the car. The shadows did not affect
the detection of the algorithm and it performed its output correctly.
Fig. 8. Successful non-detection of a gun
Finally, our last example is shown in Fig. 9. In this example, the clouds and
background did not affect the performance of the non-gun identification. The glasses
and shadows in the pictures were also not a problem with the non-gun identifications.
Fig. 9. Successful non-detection of a gun
5 Conclusions and Future Work
By looking at the presented photos we have successfully re-trained an existing

MobileNet Neural Network of varying granularity.
The assumption that using key points alone would allow testing fewer places in an
image proved to be inaccurate and was problematic when a test picture had more key
points than the maximum amount defined for detection or when there were many key
points that were detected in the same place. It was observed that in some cases so many
key points were detected on non-gun objects that the gun never got assigned a key
point and thus a gun was never classified because that area in the image was not
evaluated. It was also observed that when many key points were detected in the same
place on a gun that the classification was ran on the same area of the picture for
multiple overlapping classifications of gun.
One way to work around this issue would be to increase the maximum number of
key points, however without any mitigation this would cause the total number of places
in the picture needing to be looked at for classification to increase and remove the
benefit of using the key point detection to decrease the instances of classification versus
a sliding window over the whole picture.
Adding a processing step after the key point detection to remove key points by
using a geometric density algorithm would allow keeping the number of classifications
to a minimum. For example, in our test pictures, many of our key points are close
enough that any gun detection boxes are overlapping.
Using a second step to remove key points would benefit both busy test pictures
where the maximum number of key points was preventing a gun object from getting
any key points and non-busy test pictures where all the key points are close to each
other.
Adding another pre-processing step to remove the backgrounds would greatly
increase the gun detection and decrease the false-positives. Finally, obtaining more
pictures and continually training the dataset will allow it to continually learn and be
more accurate.
To conclude, we have presented an early gun identification algorithm that we
believe can be improved upon to allow detection of guns in images, video, and other
applications. The goal is to give security personnel extra time to respond to a real gun
alert.
References
1. Grega, M., Matiolański, A., Guzik, P., Leszczuk, M.: Automated detection of firearms and
knives in a CCTV image. Sensors 16, 1–16 (2016). https://doi.org/10.3390/s16010047.
ISSN 1424-8220
2. Tiwari, R.K., Verma, G.K.: A computer vision based framework for visual gun detection
using harris interest point detector. Procedia Comput. Sci. 54, 703–712 (2015). https://doi.
org/10.1016/j.procs.2015.06.083
3. Olmos, R., Tabik, S., Herrera, F.: Automatic Handgun Detection Alarm in Videos Using
Deep Learning. arXiv:170205147 cs (2017)
4. Yeom, S., et al.: Real-time outdoor concealed-object detection with passive millimeter wave
imaging. Opt. Express 19, 2530–2536 (2011). https://doi.org/10.1364/OE.19.002530
5. Vajhala, R., Maddineni, R., Yeruva, P.R.: Weapon Detection in Surveillance Camera Images
(2016)
6. Kang, K., Ouyang, W., Li, H., Wang, X.: Object Detection from Video Tubelets with
Convolutional Neural Networks. Presented at the June (2016)
7. Ray, L., Miao, T.: Towards Real-Time Detection, Tracking and Classification of Natural
Video. Presented at the June (2016)
8. Abadi, M.: TensorFlow: learning functions at scale. In: Proceedings of the 21st
ACM SIGPLAN International Conference on Functional Programming, p. 1. ACM, New
York (2016)
9. Angermueller, C., Pärnamaa, T., Parts, L., Stegle, O.: Deep learning for computational
biology. Mol. Syst. Biol. 12, 878 (2016)
10. Pulli, K., Baksheev, A., Kornyakov, K., Eruhimov, V.: Real-time computer vision with
OpenCV. Commun. ACM. 55, 61–69 (2012). https://doi.org/10.1145/2184319.2184337
11. Kaehler, A., Bradski, G.R.: Learning OpenCV 3: Computer Vision in C++ with the OpenCV
Library. O’Reilly Media, Sebastopol (2016)
12. Howard, A.G., et al.: MobileNets: Efficient Convolutional Neural Networks for Mobile
Vision Applications. arXiv:170404861 cs (2017)
13. Krizhevsky, A., Sutskever, I., Hinton, G.E.: ImageNet classification with deep convolutional
neural networks. Commun. ACM 60, 84–90 (2017). https://doi.org/10.1145/3065386
14. Kadiyala, A., Kumar, A.: Applications of Python to evaluate environmental data science
problems. Environ. Prog. Sustain. Energy 36, 1580 (2017)
15. Saha, M.D., Darji, M.K., Patel, N., Thakore, D.: Implementation of image enhancement
algorithms and recursive ray tracing using CUDA. Procedia Comput. Sci. 79, 516–524
(2016). https://doi.org/10.1016/j.procs.2016.03.066
16. Shi, J., Tomasi, C.: Good Features to Track. Cornell University, Ithaca (1993)
17. Xiao, J., Hays, J., Ehinger, K.A., Oliva, A., Torralba, A.: Sun database: large-scale scene
recognition from abbey to zoo. In: 2010 IEEE Conference on Computer Vision and Pattern
Recognition (CVPR). pp. 3485–3492. IEEE (2010)
18. Internet Movie Firearm Database. http://www.imfdb.org/
19. Pixabay. https://pixabay.com
20. Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., Fei-Fei, L.: Imagenet: a large-scale
hierarchical image database. In: IEEE Conference on Computer Vision and Pattern
Recognition, CVPR 2009, pp. 248–255. IEEE (2009)
Another random document with
no related content on Scribd:
Century,” November, 1891). It was suggested by “The Luck of
Roaring Camp.” The question was raised in conversation whether a
limp and molluscous baby, unable so much as to hold up its head on
its helpless little neck, could do anything so positive as to “rastle
with” Kentuck’s finger; and the more knowing persons present
insisted that a young baby does, as a matter of fact, have a good
firm hand-clasp. It occurred to Dr. Robinson that if this was true it
was a beautiful Darwinian point, for clinging and swinging by the
arms would naturally have been a specialty with our ancestors if they
ever lived a monkey-like life in the trees. The baby that could cling
best to its mother as she used hands, feet, and tail to flee in the best
time over the trees, or to get at the more inaccessible fruits and eggs
in time of scarcity, would be the baby that lived to bequeath his traits
to his descendants; so that to this day our housed and cradled
human babies would keep in their clinging powers a reminiscence of
our wild treetop days.
Dr. Robinson was fortunate enough to be able to test his theory on
some sixty babies in the first hours of their life, and was triumphantly
successful. He clasped their hands about a slender rod, and they
swung from it like athletes, without apparent discomfort, by the half
minute; many of us grown people could not do as well. Such a
remarkable power of hands and arms has for ages been of no
especial use to the human race, and it fades out in a few weeks, but
for many months the arms keep ahead of the legs in development.
Here was not only strength of arms, but the ability to perform quite
skillfully an action, that required the working together of a number of
muscles to a definite end,—the action namely, of clasping an object
with the hand. This is one of several actions that come ready-made
to the baby at birth, before he can possibly have had any chance to
learn them, or any idea of what they are for. Babies sneeze, swallow,
and cry on the first day; they shut their eyes at a bright light, or at a
touch. On the first day, moreover, they have been seen to start at a
sound or a jar; Preyer observed hiccoughing, choking, coughing, and
spreading the toes when the soles were tickled; and Darwin saw
yawning and stretching within the first week, though I do not know
that any one has seen it on the first day.
These movements are all of the class called reflex,—movements,
that is, in which the bodily mechanism is set off by some outside
action on the senses, as a gun is set off by a touch on the trigger.
Thus, when a tickling affects the mucous membrane, a sneeze
executes itself without any will of ours; when our sense of sight
perceives a swift missile coming, the neck muscles mechanically jerk
the head to one side.
We grown people have, however, a good deal of power of holding in
our reflexes,—“inhibiting” them, as the technical expression is,—but
the baby has none at all. If they had a highly developed reflex
activity, babies would be in real danger from the unrestrained acts of
their own muscles, as we see in the case of convulsions, which show
reflex action at its extreme. But the actions I have mentioned are
about all the reflex movements that have been noted in new-born
babies, except what are called the periodic reflexes, such as
breathing, the heartbeat, the contractions of the arteries, and all the
regular muscular actions of organic life.
That so complex a system of movement as these periodic reflexes
should be so readily touched into motion upon contact with air and
food, to maintain itself afterward by the interplay of the bodily
mechanism and external forces, shows a ready-made hereditary
activity far more than the sudden reflexes do. It does not work quite
smoothly at first, however: the establishment of breathing, for
instance, is irregular, and often difficult. Even the sudden reflexes
are slower and less perfect than with older people.
There is another class of movements, often confused with the reflex
—that is, instinctive movements. Real grasping (as distinguished
from reflex grasping), biting, standing, walking, are examples of this
class. They are race movements, the habits of the species to which
the animal belongs, and every normal member of the species is
bound to come to them; yet they are not so fixed in the bodily
mechanism as the reflex movements. The stimulus to them seems to
come more from within than from without—yet not from reason and
will, but from some blind impulse. This impulse is usually imperfect,
and the child has to work his own way to the mastery of the
movements. Yet though certain reflex activities are inherited in a
more highly developed condition than any human instincts, the
instincts are at bottom always hereditary, which is not the case with
the reflexes—any one may teach his muscles new reflex
movements, unknown to his ancestors. A musician does it every time
that he practices new music till his hands will run it off of their own
accord, while he is thinking of something else. But instinct cannot be
thus acquired.
The amazing instincts of the lower animals; the imperfect and broken
condition of the instincts in man, yet the deep hold that they have on
him; the mingling of inherited necessity and individual freedom in the
way in which they are worked out; the mystery of the physiological
method by which they act (while that of reflex movement is fairly well
understood, up to a certain point); the light they seem always about
to shed for the biologist on the profoundest problems of heredity, and
for the philosopher on those of free will and personality,—these
things make instinct one of the great fields of present research, and I
must not venture into it, though it is of importance in trying to
understand a baby.
I shall say only that while instinct does not appear in the lowest
animals (whose action is all of the reflex type), and is for a time a
sign of rising rank in the scale of life, it reaches its culmination with
the insects, and as we approach man it is the breaking up of the
instincts that is in its turn a sign of advancement to higher life. The
little chicken runs about as soon as it is out of its shell, and even the
monkey baby is able to take care of itself in a few months. Nothing is
so helpless as the human baby, and in that helplessness is our glory,
for it means that the activities of the race (as John Fiske has so
clearly shown) have become too many, too complex, too infrequently
repeated, to become fixed in the nervous structure before birth;
hence the long period after birth before the child comes to full human
powers. It is a maxim of biology (as well as the frequent lesson of
common observation) that while an organism is thus immature and
plastic, it may learn, it may change, it may rise to higher
development; and thus to infancy we owe the rank of the human
race.
The one instinct the human baby always brings into the world
already developed is half a mere reflex act—that of sucking. It is
started as a reflex would be, by the touch of some object, pencil,
finger, or nipple, it may be, between the lips; but it does not act like a
reflex after that. It continues and ceases without reference to this
external stimulus, and a little later often begins without it, or fails to
begin when the stimulus is given. If it has originally a reflex
character, that character fades out, and leaves it a pure instinct.
These two types of automatic movement (for instinct, however
complicated later with volition, gives rise in these earliest days to
none but automatic movement) are both “purposive,” though not
purposed—that is, they are actions that are plainly adapted to some
end by ancestral intelligence or by natural selection. But there was
another type of movements more conspicuous in our baby than
either, and apparently quite nonpurposive. From the first day she
moved slightly, but almost constantly, the legs drawing up, the arms
stirring, the eyes and head rolling a little. Sometimes the features
were distorted with vague and meaningless grimaces. Most other
observers report these movements, and inexperienced ones say that
the baby “felt with his hands about his face,” or “tried to get his
hands to his head.” Any mother may convince herself that the baby
has no will in the matter by watching till he really does begin to try,
weeks later, to turn his head, put his hands to his mouth, kick up his
legs: the difference in the whole manner of the action is evident.
An odd explanation has been offered for these movements by Dr.
Mumford, an English physiologist. He holds that they have a singular
resemblance to those of swimming amphibians; that their prototype
may be seen in any aquarium; they are, in short, survivals of the
period long before the ape-like stage, long before any mammalian
stage, when our ancestors had not yet abandoned life in the waters.
Now, although it is quite true that biologists believe that if our
ancestry is traced far enough, it does lead back to the water, still it
seems hardly possible that in a human baby, whose structure
passed the amphibian stage long before birth, the most frequent
movements should hark back to that tremendous antiquity. It is more
likely that Preyer’s explanation is the correct one: viz., that the
movements are simply due to the rapid growth of nerve centres,
which causes an overflow of nervous force to the muscles and
makes them contract at haphazard. A certain regularity is given to
these chance movements by the tendency of nerve impulse to flow
in the same paths where it has flowed before, rather than in new
ones, so that the muscles are drawn toward the position they
occupied before birth. This brings the hands constantly up about the
head—a fact that later has important results in development.
These aimless movements are called “impulsive” by Preyer. I have
followed Bain and Mrs. Moore in calling them “spontaneous.”
There were no movements beyond these three types, and therefore
none that showed the least volition. Mothers often think the crying
shows wish, will, or understanding of some sort. But Preyer tells us
that babies born without a brain cry in just the same manner.
Mothers do not like to think that the baby is at first an automaton;
and they would be quite right in objecting if that meant that he was a
mere machine. He is an automaton in the sense that he has
practically neither thought, wish, nor will; but he is a living, conscious
automaton, and that makes all the difference in the world. And it
would be a bold psychologist who should try to say what germ of
thought and will lies enfolded in his helplessness. Certainly, the
capacity of developing will is there, and an automaton with such a
capacity is a more wonderful creature than the wise, thinking, willing
baby of nursery tradition would be.
If mothers would only reflect how little developed a baby’s mind is at
a year old, after all the progress of twelve months, they would see
that they rate the mental starting point altogether too high. And they
miss thus the whole drama of the swift and lovely unfolding of the
soul from its invisible germ—a drama that sometimes fairly catches
one’s breath in the throat with excitement and wonder.
III
THE NEW-BORN BABY: SENSATIONS AND
CONSCIOUSNESS.
I have said that the baby began the world as an automaton, but a
conscious, feeling automaton. And what, then, were these feelings
and this consciousness? What was the outfit for beginning the world
that the little mind brought with it? When I asked such questions I
was skirting the edge of one of the great battle-grounds of
philosophy. Whether all human ideas are made up solely from one’s
own experience of the outer world as given him by his senses, or
whether there are, on the contrary, inborn ideas, implanted directly
by nature or God,—this is a question on which volumes have been
written.
Did the baby start out ready equipped with ideas of space, personal
identity, time, causation, such as we find so ineradicable in our own
minds? That is, did she see objects about her, located in space,
nearer and farther, right and left, and all outside and separate from
herself, as we do? hear sounds coming from without, as we do? Did
she feel herself a separate thing from the outer world? Did she
perceive events as happening in time succession, one after another?
And did she think of one thing as happening because of another, so
that, for instance, she was capable of crying in order to cause her
dinner to be brought?
The hope of answering such questions was the first stimulus to the
study of infants, and the earlier records are much occupied with
them. Philosophers nowadays are less disposed to think that we can
prove anything about the doctrine of innate ideas by finding whether
babies have such ideas to begin with; for we might indeed have
ideas that came direct from God, or from the nature of the mind, and
yet might not enter into our inheritance of these at once.
To me, however, not seeking to solve philosophical problems, but
only to watch and comprehend what was going on in the baby’s
mind, it was none the less interesting to try to make out the condition
of her senses and consciousness—though without the careful
special investigations certain physiologists had made before, I
should have found it blind guessing as to how much she really did
see, hear, and feel; for these processes, of course, went on inside
her little mind, and could only be inferred from her behavior.
She evidently felt a difference between light and darkness from the
first hour, for she stopped crying when her face was exposed to
gentle light; and other observers confirm this. Two or three report
also a turning of the head toward the light within the first week. The
nurse, who was intelligent and exact, thought she saw this in the
case of my niece. I did not, but I saw instead a constant turning of
the eyes toward a person coming near her—that is, toward a large
dark mass that interrupted the light. Either movement must be
regarded as entirely instinctive or reflex. Even plants will turn toward
the light, and among animal movements this is one of the most
primitive; while the habit of looking toward any dark moving mass
runs far back in animal history, and may well have become fixed in
the bodily mechanism. With the beginning of voluntary looking these
instinctive movements fade.
No other sign of vision appeared in the little one during the first
fortnight. The eyes were directed to nothing, fixed on nothing. They
did not wink if one made a pass at them. There was no change of
focus for near or distant seeing; the two eyes did not even move
always in unison,—and as the lids also had by no means learned yet
to move symmetrically with the balls and with each other, some
extraordinary and alarming contortions resulted.
True seeing, such as we ourselves have, is not just a matter of
opening the eyes and letting the vision pour in; it requires a great
deal of minute muscular adjustment, both of the eyeballs and of the
lenses, and it is impossible that a baby should see anything but blurs
of light and dark (without even any distinction of distance) till he has
learned the adjustments. Not colored blurs, but light and dark only,
for no trace of color sense has ever been detected within the first
fortnight of life, no certain evidence of it even within the first year.
The baby showed no sign of hearing anything until the third day,
when she started violently at the sound of tearing paper, some eight
feet from her. After that, occasional harsh or sudden sounds—
oftener the rustling of paper than anything else—could make her
start or cry.
It is well established by the careful tests of several physiologists that
babies are deaf for a period lasting from several hours to several
days after birth. The outer tube of the ear is often closed by its own
walls, and the middle ear is always stopped up with fluid. Even after
the ear itself is clear and ready for hearing, few sounds are noticed;
perhaps because the outer passage is still so narrow, perhaps
because of imperfect nerve connections with the brain, perhaps
because sounds are not distinguished, but go all together into a sort
of blur, just as the sights do. As the usual effect of sounds on wee
babies is to startle them, and to set off convulsive reflex movements,
it is well for them that hearing is so tardy in development.
There is noticeable variation in sensitiveness to hearing, not only
among different babies, but in the same baby at different times. A
sound that startles on one day seems to pass absolutely unheard on
the next.
In observing the sensibility to sound, one may easily be misled. If a
baby starts when a door slams or a heavy object falls, it is more
likely to be the jar than the sound that affects him; if he becomes
restless when one claps the hands or speaks, it may be because he
felt a puff of air on his head. The tap of an ordinary call bell is a good
sound to test with, causing neither jar nor air current.
Taste and smell were senses that the baby gave no sign of owning
till much later. The satisfaction of hunger was quite enough to
account for the contentment she showed in nursing; and when she
was not hungry she would suck the most tasteless object as
cheerfully as any other. Physiologists, however, have had the daring
to make careful test of smell and taste in the new-born, putting a wee
drop of quinine, sugar, salt, or acid solution on the babies’ tongues,
and strong odors to their noses, and have been made certain by the
resulting behavior that these senses do exist from the first. But it
requires rather strong tests to call them into action. Many babies, for
instance, suck at a two per cent. solution of quinine as if it were
sugar; so it seems unlikely that the mild and monotonous taste of
milk, and the neutral smells by which any well-kept baby is
surrounded, are really perceived at all. There are instances related
of very positive discrimination between one milk and another, either
by taste or smell, shown by very young babies; yet the weight of
evidence points to an almost dormant condition of these two senses.
We were told in school that the fifth sense was “feeling,” but
psychologists now regard this not as a single sense, but as a group,
called the “dermal” or skin senses. The sense of touch and pressure,
the senses of heat and cold, and the sense of pain are the principal
ones of the group.
Our baby showed from the first that she was aware when she was
touched. She stopped crying when she was cuddled or patted. She
showed comfort in the bath, which may have been in part due to
freedom from the contact of clothes, and to liking for the soft touches
of the water. She responded with sucking motions to the first touch of
the nipple on her lips. Preyer found the lips of new-born babies quite
delicately sensitive, responding even to the lightest touch; and there
are other sensitive spots, such as the nostrils and the soles of the
feet.
On the whole, however, the rose-leaf baby skin proves to be much
less sensitive than ours, not only to contact, but also to pain, and
perhaps to heat and cold, though this has not been so thoroughly
tested. This is not saying, of course, that the physiological effects of
heat and cold upon the baby are unimportant.
Our baby had no experience of skin pain in her early days, and being
kept at an equable temperature, probably received no definite
sensations either of heat or of cold.
The foregoing are the “special senses,” that is, those that give
impressions of external things, and have end organs to receive and
make definite these impressions,—the eye at the end of the optic
nerve, the different kinds of nerve tips in the skin, and so forth.
Another sense now claims almost to rank with them,—the recently
studied sense of equilibrium and motion, by which we feel loss of
balance in our bodies and changes in their motion (changes only, for
no one can feel perfectly smooth motion). This sense has been
traced to the semicircular canals of the ear; and as this part of the
ear is the oldest in evolution, and the rudimentary ears of the lower
orders of animals are quite analogous to it in structure, biologists
now suspect that hearing may be a more recent sense than we have
thought, and that much which has been taken for sense of sound in
the lower animals—even as high as fishes—may perhaps be only a
delicate sense of motion.
I failed to watch for this motion sense in the baby. It would have been
shown by signs that she felt change of motion when she was lifted
and moved. Equilibrium sense she must have used as soon as she
began to balance her little head, but in the first limp and passive
days there was no sign of it. Still, there are tales of very young
babies who showed disturbance, as if from a feeling of lost
equilibrium, when they were lowered swiftly in the arms.
There is besides a sort of sensibility to vibration that affects the
whole body. We know how much of the rhythm of music may be
caught quite soundlessly through, the vibrations of the floor; and it is
said (perhaps not altogether credibly) that it was thus that Jessie
Brown recognized even the instruments and the tune at the relief of
Lucknow by the tremor along the ground before a sound was
audible. A jar, affecting the whole body, seems to be felt by creatures
of very low organization. Babies are undoubtedly quite susceptible to
jarring from the earliest days. Champney’s baby started when the
scale of the balance in which he was lying immediately after birth
sprang up.
Then there is the “muscle sense”—the feeling of the action of our
own muscles; and a most delicate and important sense this is. It is
safe to say that the baby had it from the first, and felt the involuntary
movements her own little body was making, for it is hardly
conceivable how else she could have learned to make voluntary
ones. But that is another story, and comes later.
Even this does not exhaust the list of sensations the baby could feel.
There was the whole group of “organic sensations,” coming from the
inner organs,—hunger, thirst, organic pain. With older people,
nausea, suffocation, choking, and perhaps some others might be
added; but little babies certainly do not feel nausea,—their food
regurgitates without a qualm. Nor do they seem to feel disagreeable
sensations when they choke in nursing.
Organic pain our baby had her touch of in the usual form of colic;
and hunger was obviously present very early, though perhaps not in
the first two or three days. Thirst appeared from the first, and was
always imperative. Of course, the milk diet largely satisfied it, but not
entirely. Luckily our baby did not suffer from thirst, for grandma,
nurse, and the good doctor had all entered early warning that
“babies needed water,” and that many a baby was treated for colic,
insomnia, nervousness, and natural depravity, when all the poor little
fellow wanted was a spoonful of cool water. The baby’s body, as I
said in my last chapter, is largely composed of water, and the
evaporation from the loose texture of the skin is very great. After
children can talk, they wear out the most robust patience with
incessant appeals, night and day, for a “d’ink,” and consume water in
quantities quite beyond what seems rational. But their craving is
doubtless a true indication of what they need.
There are composites of sensation which the baby experiences very
early. There is the feeling of clothes, for instance, made up of
warmth, of touch and pressure sensations all over his skin, and of
changes in the muscular feelings from constraint, and in the internal
feelings from the effect on circulation. There are feelings of fatigue in
one position, made up of sensations of touch, of the pressure of the
body’s weight on the under surfaces of skin, of some muscular
tensions, and perhaps of several other elements. Our baby’s nurse
saved her much fretting by simply changing the position of the little
body from time to time. We ourselves are constantly moving and
shifting our positions, to relieve a pressure on the skin here, or a
muscular tension there, but the wee baby cannot so much as turn his
head or move a limb at will.
Vaguest and most composite of all is what is called “common
sensation,” or “general sensation”—that feeling of comfort or
discomfort, vigor or languor, diffused through the whole body, with
which we are all familiar. It seems to be very primitive in origin—
indeed, the speculation is that this dim, pervasive feeling is the
original one, the primitive way in which animal tissue responded to
light and heat and everything, before the special senses developed,
gathering the light sensations to one focus, the sound sensations to
another, and so on. But in its present development it is also largely
made up of the sum of all the organic sensations, and even of dim
overflows of feeling from the special senses.
It is with older people notably connected with emotional states. It
varies, of course, with health and external conditions; yet each
person seems from birth to be held to a certain fixed habit in this
complex underlying condition of feeling—pleasant with one,
unpleasant with another. This fixed habit of general sensation is
perhaps the secret of what we call temperament; while its surface
variations seem to be mainly responsible for moods.
Our baby showed temperament—luckily of the easy-going and
cheerful kind—from her first day (though we could hardly see this
except by looking back afterward); and there is no reason to doubt
that she experienced some general sensation from the first. It was
evidently of a pretty neutral sort, however: the definite appearance of
high comfort and well-being did not come till later; nor were moods
apparent at first.
Now in all this one significant thing appears. Sensations had from
the first the quality of being agreeable or disagreeable. The baby
could not wish, prefer, and choose, for she had not learned to
remember and compare; but she could like and dislike. And this was
shown plainly from the first hour by expressions of face—reflex facial
movements, so firmly associated in the human race with liking and
disliking that the most inexperienced observer recognizes their
meaning at once. It is said that facial expression comes by imitation,
and that the blind are therefore deficient in it; but this is not true of
these simplest expressions: they come by inheritance, and are
present in the first hour of life. A look of content or discontent, the
monotonous cry, and vague movements of limbs, head, and
features,—these are the limits of expression of feeling in the earliest
days.
It would seem that in this sense condition there was nothing that
could give the baby any feeling of inner or outer, of space or locality.
We have some glimpse of the like condition ourselves,—when
people say after an explosion, for instance, that it “seemed to be
inside their own heads,” or when we try to locate a cicada’s note, or
when we feel diffused warmth.
Here is the conception I gathered of the dim life on which the little
creature entered at birth. She took in with a dull comfort the gentle
light that fell on her eyes, seeing without any sort of attention or
comprehension the moving blurs of darkness that varied it. She felt
motions and changes; she felt the action of her own muscles; and,
after the first three or four days, disagreeable shocks of sound now
and then broke through the silence, or perhaps through an unnoticed
jumble of faint noises. She felt touches on her body from time to
time, but without the least sense of the place of the touch (this
became evident enough later, as I shall relate in its order); and
steady slight sensations of touch from her clothes, from arms that
held her, from cushions on which she lay, poured in on her.
From time to time sensations of hunger, thirst, and once or twice of
pain, made themselves felt through all the others, and mounted till
they became distressing; from time to time a feeling of heightened
comfort flowed over her, as hunger and thirst were satisfied, or
release from clothes, and the effect of the bath and rubbing on her
circulation, increased the net sense of well-being. She felt slight and
unlocated discomforts from fatigue in one position, quickly relieved
by the watchful nurse. For the rest, she lay empty-minded, neither
consciously comfortable nor uncomfortable, yet on the whole
pervaded with a dull sense of well-being. Of the people about her, of
her mother’s face, of her own existence, of desire or fear, she knew
nothing.
Yet this dim dream was flecked all through with the beginnings of
later comparison and choice. The light was varied with dark; the
feelings of passive motion, of muscular action, of touch, of sound,
were all unlike each other; the discomforts of hunger, of pain, of
fatigue, were different discomforts. The baby began from the first
moment to accumulate varied experience, which before long would
waken attention, interest, discrimination, and vivid life.
IV
THE EARLIEST DEVELOPMENTS
Out of the new-born baby’s dim life of passivity the first path was
that of vision. I noticed about the end of the second week that her
eyes no longer wandered altogether helplessly, but rested with a
long and contented gaze on bright surfaces they chanced to
encounter, such as the shining of the lamp on the white ceiling, or
our faces turned toward the light as she lay on our knees. It was not
active looking, with any power to direct the eyes, but mere staring;
when the gaze fell by chance on the pleasant light, it clung there. But
something must have come to pass, that it could stop and cling to
what gave it pleasure.
I think no one has yet analyzed this earliest stage in progress toward
real seeing, though Professor Sully touches on an explanation when
he says that the eyes “maintain their attitude under stimulus of the
pleasure.”
We know that muscular action is normally caused by stimulus
received from the nerve centres, and that in the earliest days there
seems to be a good deal of random discharge of stimulus,
developed by the growth of the centres, and causing aimless
movements. Now there are two fundamental and profoundly
important things about this nervous discharge. One is that pleasure,
attention, or intensity of sensation seems to have the power of
increasing it, and thus influencing the action of the muscles. The
other is that the discharge always tends to seek the same paths it
has used before, and more and more easily each time; so that
physiologists speak of it as a current deepening its channels. It is
really nothing like a flowing liquid, nor the nerve threads along which
it passes like channeled watercourses. Still, just as a current of water
will deepen a gully till it drains into itself all the water that had spread
about in shallower ditches, so the wave of molecular change running
along a nerve somehow so prepares that nerve that by and by,
instead of spreading about through any fibres that come handy, the
whole energy will drain into the accustomed ones. Then, of course,
the muscles to which these run will perform more and more easily
the accustomed acts. Some of these channels—even whole
connected systems of them—are already well prepared by
inheritance, and hence come instinctive and reflex actions; many are
still to be deepened by the baby’s own experience.
Now suppose the aimless impulse straying to the baby’s eye
muscles, making the eyes roam hither and yon; but as they reach a
certain position, they fall upon a lighted surface, and a pleasant
brightness flows back into the consciousness; and something stirs
within that has power to send an intenser current through those
same fibres. For the time, at least, that channel is deepened, the
wandering impulses are drained into it, and the eye muscles are held
steady in that position. And, in fact, with the beginning of staring the
irregular movements of head and eyes did decline, and gradually
disappear.
It is an important moment that marks the beginning of even a
passive power to control the movements; and when my grandmother
handed down the rule that you should never needlessly interrupt a
baby’s staring, lest you hinder the development of power of attention,
she seems to have been psychologically sound.
A fuller and pleasanter life now seemed to pervade the whole little
body. The grimaces of vague discomfort were disappearing, and the
baby began to wear a look of satisfaction as she lay, warm and fed
and dry, gazing at some light surface. In the bath, where the release
from clothes and the stimulus to circulation from the warm water
heightened the pleasant condition of general sensation, her
expression approached real delight; the movements of her limbs
were freer, and all her muscles tenser.
The neck muscles, especially, were so far “innervated”—that is,
supplied with nervous energy—as fairly to lift her head from the
supporting hand. This was probably not as yet a real effort to hold up
the head, only a drafting of surplus energy into the neck muscles,
partly because of inherited aptitude, partly because the pleasure
received from the lifted head and better seeing tended to draw the
energy thither, just as it was drawn to the eye muscles in the case of
the staring. At least one careful observer, Mrs. Edith Elmer Wood,
records this action of the neck muscles on the first day.
It was at this period that the baby first smiled; but being forewarned
of the “colic smile,” which counterfeits so exactly the earliest true
smiles,—fleeting as these are, just touching the mouth and
vanishing,—I never felt sure whether the baby was smiling for
general contentment with life, or whether a passing twinge had
crossed her comfort and drawn her lips into the semblance of a
smile; and so never dared to record the expression till it first
occurred for unmistakable pleasure.
There must have been rapid progress going on in the clearness of
muscular and touch sensations, and in the forming of associations in
the baby’s mind; but no plain evidence of these inner processes
came till the fourth week. Then I noticed that the baby, when crying
with hunger, would hush as soon as she was taken in the arms in the
position usual in nursing, as if she recognized the preliminaries, and
knew she was about to be satisfied. She could not, in fact, have
remembered or expected anything as yet; it was not memory, but a
clear instance of the working of that great law of association by
which the raw material of the senses was to be wrought up into an
orderly mental life.
The substance of the law is that when experiences have repeatedly
been had together, the occurrence of one of them (still more, of
several out of a group, as in this case) tends to bring up into
consciousness the others. It is a law that underlies psychic life as
profoundly as the law that nerve energy seeks its old channels
underlies physical life. Indeed, it is in a sense the psychic side of the
same law; for it implies that when a group of nerve centres have
formerly acted together, the action of one tends to bring on that of
the rest. So, since the baby had often experienced the feeling of that
particular position (a combination of tactile and muscular and organic
sensations) in connection with the feeling of satisfied hunger, that
comfortable feeling, the missing member of the group, came into her
consciousness along with the rest, some moments in advance of the
actual satisfaction.
I have said that this is not memory, yet there is in it a germ of
memory. A past experience is brought back to consciousness; and if
it were brought back as a definite idea, instead of a vague feeling, it
would be memory.
Close on this came another great advance in vision. This was on the
twenty-fifth day, toward evening, when the baby was lying on her
grandmother’s knee by the fire, in a condition of high well-being and
content, gazing at her grandmother’s face with an expression of
attention. I came and sat down close by, leaning over the baby, so
that my face must have come within the indirect range of her vision.
At that she turned her eyes to my face and gazed at it with the same
appearance of attention, and even of some effort, shown by a slight
tension of brows and lips, then turned her eyes back to her
grandmother’s face, and again to mine, and so several times. The
last time she seemed to catch sight of my shoulder, on which a high
light struck from the lamp, and not only moved her eyes, but threw
her head far back to see it better, and gazed for some time, with a
new expression on her face—“a sort of dim and rudimentary
eagerness,” says my note. She no longer stared, but really looked.
Clear seeing, let us here recall, is not done with the whole retina, but
only with a tiny spot in the centre, the so-called “yellow spot,” or
“macula lutea.” If the image of an object falls to one side of this,
especially if it is far to one side, we get only a shapeless impression
that something is there; we “catch a glimpse of it,” as we say. In
order really to look at it we turn our eyeballs toward the object till the
image falls on the spot of clear vision. We estimate the distance
through which to turn the balls, down to minute fractions of an inch,
by the feeling in the eye muscles.
This was what the baby had done, and I do not dare to say how
many philosophical and psychological discussions are involved in
her doing it. Professor Le Conte thinks that it shows an inborn sense
of direction, since the eyes are turned, not toward the side on which
the ray strikes the retina, but toward the side from which the ray
enters the eye; that is, the baby thinks out along the line of the ray to
the object it comes from, thus putting the object outside himself, in
space, as we do. Professor Wundt, the great German psychologist,
is positive that the baby has no sense of space or direction, but
gains it by just such measurements with the eye muscles; that there
is no right nor left, up nor down, for him, but only associations
between the look of things off at one side, and the feel of the eye
action that brings them to central vision.
This means that before a baby can carry the eye always through just
the right arc to look at an object, he must have made this association
between the look of things and the feel of the action separately for
each point of the retina. It is a great deal for a baby to have learned
in three weeks; still, babies have to learn fast if they are ever to
catch up with the race; and in the early roamings of the eye they
experience over and over all manner of transits of images to and fro
across the retina. Probably, too, it was still only partially learned.
I watched now for what Preyer’s record had led me to expect as the
next development in vision—the ability to follow a moving object with
the eyes; that is, to hold the yellow spot fixed on the object as it
moved, moving the eyeball in time with it in order to do so. I used my
hand to move to and fro before the baby, and could not satisfy
myself that she followed it, though she sometimes seemed to; but
the day after she was a month old I tried a candle, and her eyes
followed it unmistakably; she even threw her head back to follow it
farther. In trying this experiment, one should always use a bright
object, should make sure the baby’s eyes are fixed on it, and then
should move it very slowly indeed, right and left.
So far, there is no necessary proof of will. Longet found that the eyes
and head of a pigeon whose cerebrum had been removed would
follow a moving light. We ourselves can sit absorbed in thought or
talk, yet follow unconsciously with our eyes the movement of a
lantern along a dark road; and if something appears on the outer
edge of our vision we often turn quite involuntarily to look. But the
baby’s new expression of intelligence and interest showed that
whether she willed the movements or not, she attended to the new
impressions she was getting.
Professor Preyer noticed the same dawn of intelligence in his baby’s
face at about the same stage. And it is worth while to observe that
when I came to study my record I was surprised to find how often
such an awakening look, an access of attention, wonder, or
intelligence, in the baby’s face, had coincided with some marked
step in development and signalized its great mental importance. I
should advise any one who is observing a baby to be on the lookout
for this outward and visible sign of an inward and spiritual unfolding.
In both these visual developments the baby had proved able to use
her neck in coöperation with her eyes, throwing back her head to see
farther. It began at the same time to seem that she was really and
deliberately trying to hold up her head for the same purpose of
seeing better. She not only straightened it up more and more in the
bath, but when she was laid against one’s breast she would lift her
head from the shoulder, sometimes for twenty seconds at a time,
and look about. Preyer sets this down as the first real act of will.
The baby’s increased interest in seeing centred especially on the
faces about her, at which she gazed with rapt interest. Even during
the period of mere staring, faces had oftenest held her eyes,
probably because they were oftener brought within the range of her
clearest seeing than other light surfaces. The large, light, moving
patch of the human face (as Preyer has pointed out) coming and
going in the field of vision, and oftener chancing to hover at the point
of clearest seeing than any other object, embellished with a play of
high lights on cheeks, teeth, and eyes, is calculated to excite the
highest degree of attention a baby is capable of at a month old. So
from the very first—before the baby has yet really seen his mother—
her face and that of his other nearest friends become the most active
agents in his development, and the most interesting things in his
experience.
Our baby was at this time in a way aware of the difference between
companionship and solitude. In the latter days of the first month she
would lie contentedly in the room with people near by, but would fret
if left alone. But by the end of the month she was apt to fret when
she was laid down on a chair or lounge, and to become content only
when taken into the lap. This was not yet distinct memory and

Ebffiledoc 3819

Uploaded by

Copyright:

Available Formats

Ebffiledoc 3819

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ebffiledoc 3819

Uploaded by

Copyright:

Available Formats

Machine Learning and Intelligent

Communications Third International

Machine Learning and Intelligent Communications Second

Service Oriented Computing 16th International

Intelligent Information Processing X 11th IFIP TC 12

Machine Learning and Intelligent Communications First

Web Information Systems and Applications: 15th

Intelligent Systems Design and Applications: 18th

Intelligent Systems Design and Applications: 18th

ICCCE 2018: Proceedings of the International Conference

ISSN 1867-8211 ISSN 1867-822X (electronic)

Library of Congress Control Number: 2018954076

August 2018 Limin Meng

Steering Committee Co-chair

Steering Committee Member

Technical Program Committee Chairs

Publicity and Social Media Chair

Sponsorship and Exhibits Chair

Technical Program Committee

Intelligent Positioning and Navigation

Intelligent Multimedia Processing and Security

Wireless Mobile Network and Security

Cognitive Radio and Intelligent Networking

Intelligent Internet of Things

Intelligent Satellite Communications and Networking

Intelligent Remote Sensing, Visual Computing and Three-Dimensional Modeling

Green Communication and Intelligent Networking

Intelligent Ad-hoc and Sensor Networks

Intelligent Resource Allocation in Wireless and Cloud Networks

Intelligent Cooperative Communications and Networking

Intelligent Computing and Caching

Intelligent Signal Processing in Wireless and Optical Communications

Intelligent Radar Signal Processing

Gun Identification Using Tensorflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

A Research of Network Applications Classification Based

Real-Time Drone Detection Using Deep Learning Approach . . . . . . . . . . . . 22

Deep Reinforcement Learning-Based Task Offloading and Resource

RFID Data-Driven Vehicle Speed Prediction Using Adaptive

Speed Prediction of High Speed Mobile Vehicle Based on Extended

Mobility Prediction Based on POI-Clustered Data . . . . . . . . . . . . . . . . . . . . 60

A Research on the Identification of Internet User Based on Deep

Intelligent Positioning and Navigation

An Improved Constrained Least Squares Localization Algorithm in NLOS

Intelligent Multimedia Processing and Security

Fast Inter Prediction Mode Decision Algorithm Based on Data Mining . . . . . 95

Wireless Mobile Network and Security

Smartphone Application Identification by Convolutional Neural Network . . . . 105

Joint D2D Cooperative Relaying and Friendly Jamming Selection

Variable Tap-Length Blind Equalization for Underwater Acoustic

Random Sequence Generation Algorithm for Multi-chaotic Systems . . . . . . . 136

VulAware: Towards Massive-Scale Vulnerability Detection in Cyberspace . . . 144

Cognitive Radio and Intelligent Networking

Optimization in Cognitive Radio Networks with SWIPT-Based DF Relay . . . 153

Real-Valued Orthogonal Sequences for Ultra-low Overhead Channel

A Novel Spectrum Allocation Scheme in Femtocell Networks Using

Distributed Channel Allocation for Wireless Mesh Networks

Physical Violence Detection with Movement Sensors . . . . . . . . . . . . . . . . . 190

Joint Resource Allocation for Wireless Energy Harvesting Based

Optimal Power Splitting of Cognitive Radio Networks with