In this paper, we present our recent work on cells detection and segmentation in estrogen receptor immunohistochemistry (ER-IHC)-stained breast carcinoma images. The proposed cell detection and segmentation is very useful in the... more
In this paper, we present our recent work on cells detection and segmentation in estrogen receptor immunohistochemistry (ER-IHC)-stained breast carcinoma images. The proposed cell detection and segmentation is very useful in the predictive scoring of hormone receptor status in ER-IHC stained whole-slide images, which helps pathologists to decide whether a patient should be offered hormonal therapy or other treatments. The proposed method is based on deep convolutional neural network, followed by watershed-based post-processing. The cell detection results are compared and evaluated objectively against the ground truth provided by our collaborating pathologists. The cell segmentation results, on the other hand, are evaluated visually by overlaying the computer segmented boundaries on the ER-IHC images for comparison. The automated cell detection algorithm recorded precision and recall rates of 95% and 91% respectively. The very promising performances for both the detection and segmentation paves the way for an automated system for hormone receptor scoring in ER-IHC stained whole-slide breast carcinoma images.
Research Interests:
Research Interests:
The advent of multimedia technology has offer new dimension to computerized applications. Art-based applications are among those which have and will continue to benefit from this advancement. With the ever growing size and variety of... more
The advent of multimedia technology has offer new dimension to computerized applications. Art-based applications are among those which have and will continue to benefit from this advancement. With the ever growing size and variety of accessible data across museum collections, the need for flexible and efficient data retrieval is growing at an alarming rate. Content-based image retrieval (CBIR) and analysis is getting a lot of attention from museums and art institutions. One of the image-based requirements from museums, is to automatically classify craquelure (cracks) in paintings for the purpose of aiding damage assessment. Craquelure in paintings can be an important element in judging authenticity, use of material as well as environmental and physical impact because these can lead to different craquelure patterns. Mass screening of craquelure patterns will help to establish a better platform for conservators to identify cause of damage. As a way of performing such action, a content...
Research Interests:
In this paper, we present a 2-stage approach to connected curve grouping. The algorithm is experimented and demonstrated on crack-detected images of paintings. Some features are left undetected and this tends to produce disconnected... more
In this paper, we present a 2-stage approach to connected curve grouping. The algorithm is experimented and demonstrated on crack-detected images of paintings. Some features are left undetected and this tends to produce disconnected curves. In order to extract high-level features for content-based application, these supposedly connected curves have to be grouped together. It is one of the many steps needed to produce a content-based platform for digital analysis of crack patterns in paintings particularly for classification purpose. The prime objective of the grouping algorithm is to segment or partition areas of an image to produce reliable representations of content. The first stage of the algorithm utilizes the Minimum Bounding Rectangle (MBR) of a crack network as means of finding overlapping features. We demonstrate the use of the both the rotated and the un-rotated MBR. In the second stage, curve characteristics represented by the rotated MBR such as the dimension ratio, the a...
Research Interests:
Human action recognition is one of the most challenging computer vision topics and an important realistic yet open research task in many relevant applications such as security, content retrieval and video surveillance. In this paper, we... more
Human action recognition is one of the most challenging computer vision topics and an important realistic yet open research task in many relevant applications such as security, content retrieval and video surveillance. In this paper, we proposed a robust approach for human action recognition in different video scenes. The architecture of the approach consists of shot boundary detection, shot frame rate resampling, extracting compact feature vectors by emphasizing variation and classification. Extracting strong patterns in feature vectors before classification affect the chosen approach high performance compared with latest state-of-art work using four widely used realistic data sets: Hollywood-2, UCF11(YouTube actions), UCF Sports and KTH.
In this paper, we extended our previous non-adaptive human action recognition framework by incorporating an adaptive model in order to exhibit minimal supervision system where the intelligent system will handle the process of human action... more
In this paper, we extended our previous non-adaptive human action recognition framework by incorporating an adaptive model in order to exhibit minimal supervision system where the intelligent system will handle the process of human action detection, labelling, training and recognition in automated manner. Action descriptor that is formed in non-adaptive model framework is compared with the stored action descriptors in order to determine a new human action. Then, the collected new action descriptors are clustered using agglomerative hierarchical clustering algorithm. Subsequently, all the clustered action descriptors are assigned with new labels and the multi-class SVM classifier is updated automatically. Additionally, we evaluated the adaptive model on public depth dataset namely MSR-Action 3D dataset. The experimental results reveal that the proposed adaptive framework is comparable with state of the arts methods which were developed in supervised manner. In future, conventional RG...
Recognition of human action is one of the challenges in the field of artificial intelligence. Deep learning model has become a research issue in action recognition applications due to its ability to outperform traditional machine learning... more
Recognition of human action is one of the challenges in the field of artificial intelligence. Deep learning model has become a research issue in action recognition applications due to its ability to outperform traditional machine learning approaches. The Convolutional Neural Network is one of the architectures commonly used in most action recognition works. There are different models in the Convolutional Neural Network, but no study has been done to evaluate which model has the best performance in understanding human actions. Thus, in this paper, we compare the performance of two separate pre-trained models of deep Convolutional Neural Network in classifying the human actions to identify the different behaviours. GoogleNet and AlexNet are the used two models with fine-tuned parameters used for comparison, in addition, to use Long-Short Term Memory for the video's labels prediction. The paper's main contribution is that it offers a performance analysis of two separate fine-tu...
Electrocardiogram (ECG) signal compression is playing a vital role in biomedical applications. The signal compression is meant for detection and removing the redundant information from the ECG signal. Wavelet transform methods are very... more
Electrocardiogram (ECG) signal compression is playing a vital role in biomedical applications. The signal compression is meant for detection and removing the redundant information from the ECG signal. Wavelet transform methods are very powerful tools for signal and image compression and decompression. This paper deals with the comparative study of ECG signal compression using preprocessing and without preprocessing approach on the ECG data. The performance and efficiency results are presented in terms of percent root mean square difference (PRD). Finally, the new PRD technique has been proposed for performance measurement and compared with the existing PRD technique; which has shown that proposed new PRD technique achieved minimum value of PRD with improved results.
The proposed method is a contrast measurement technique which makes use of the second-order derivatives of the pixels in MRI images. The technique utilizes the changes in the histogram of second-order derivatives for the image,... more
The proposed method is a contrast measurement technique which makes use of the second-order derivatives of the pixels in MRI images. The technique utilizes the changes in the histogram of second-order derivatives for the image, specifically the region under a Laplacian curve which approximates the distribution of the pixel frequencies in the histogram. As gain is introduced, the region of the histogram under the curve changes. The standard deviation of the frequency differences between the histogram and the curve in the region is used as a measurement score. The objective of this method is to provide a way to include small details, such as small blood vessels, in the contrast measurement, compared to local average contrast and RMS contrast measurements which do not. The outcome is a scoring method which not only provides a measurement of contrast, but also indicates the loss of small details in MRI images.
Research Interests:
Histopathological analysis of tissues has been gaining a lot of interests recently, from developing computer algorithms to assist pathologists from cell detection and counting, to tissue classification and cancer grading. With the advent... more
Histopathological analysis of tissues has been gaining a lot of interests recently, from developing computer algorithms to assist pathologists from cell detection and counting, to tissue classification and cancer grading. With the advent of whole slide imaging, the field of digital pathology has gained enormous popularity, and is currently regarded as one of the most promising avenues of diagnostic medicine. Deep learning advancement on image set today has successfully evolved as many models has been proposed and produced state-of-the-art object classifying results. This is not limited to large database such as Imagenet but also has seen applications in other medical image analysis related areas. In this paper we have carefully constructed and expanded the deep model network to classify normal and tumor slides in histology images of lymph nodes tissue. We have proposed our own deep learning model based on convolutional neural network with smaller requirement using 64×64×3 input imag...
Locally Linear Embedding (LLE) is an unsupervised non-linear manifold learning method, which has spurred increased interest in face recognition research recently. However, it is commonly known that a supervised method that considering the... more
Locally Linear Embedding (LLE) is an unsupervised non-linear manifold learning method, which has spurred increased interest in face recognition research recently. However, it is commonly known that a supervised method that considering the class-specific information always outperforms the unsupervised one, especially in biometric recognition task. In this paper, we propose a supervised LLE technique, known as class-label Locally Linear Embedding (cLLE). cLLE aims to discover the nonlinearity of high-dimensional data by minimizing the global reconstruction error of the set of all local neighbours in the data set. cLLE method is using user class-specific information in neighbourhoods selection and thus preserves the local neighbourhoods. Since the locality preservation is correlated to the class discrimination, the proposed cLLE is expected superior to LLE in face recognition. Experimental results on three face databases demonstrate the success of the proposed technique.
In sports science, two widely used approaches to perform movement recognition and analysis are through manual annotation of sports video and physical body marker attached to athlete’s body. The use of physical body markers, however,... more
In sports science, two widely used approaches to perform movement recognition and analysis are through manual annotation of sports video and physical body marker attached to athlete’s body. The use of physical body markers, however, requires expertise on visual annotation which is obviously time-consuming and inconvenient for the athletes. Badminton is one of Malaysia’s most popular sports but there is still a lack of scientific research on movement recognition and analysis focusing on this sport. Therefore, in this paper, a novel lossless compact view invariant compression technique with a dynamic time warping algorithm is proposed to cater for both badminton movement recognition and analysis frameworks. Our experimental dataset of depth map sequences composed of 10 types of badminton movements with a total of 600 samples performed by 20 badminton players. The dataset varies in terms of viewpoints, human body size, clothes, speed, and gender. Experimental results revealed that near...
With the increasing prevalence rate of diabetes and obesity worldwide, chronic wounds are becoming a significant burden for world health and economy. The treatment of a chronic wound goes through complex and time-intensive process. During... more
With the increasing prevalence rate of diabetes and obesity worldwide, chronic wounds are becoming a significant burden for world health and economy. The treatment of a chronic wound goes through complex and time-intensive process. During the healing period, continuous wound measurement helps clinicians to predict the healing time and monitor the treatment efficiency. Current clinical techniques such as ruler-based or tracing-based methods are inaccurate, time-consuming and also subject to intra- and inter-reader variability that does not satisfy a comprehensive clinical benchmark. In this paper, we proposed a method for wound boundary demarcation and estimation based on Super pixel segmentation and classification using an enhanced convolution neural network. An overall accuracy, sensitivity, and specificity of around 90% was observed, which fared much better against traditional methods.
Chronic wound is becoming a major threat for world health and economy. In the USA alone, an estimated 6.5 million people are affected by the chronic wound and the annual cost for chronic wound treatment is reportedly more than 25 billion... more
Chronic wound is becoming a major threat for world health and economy. In the USA alone, an estimated 6.5 million people are affected by the chronic wound and the annual cost for chronic wound treatment is reportedly more than 25 billion dollars. The process of chronic wound healing is very complex and time-consuming. Quantification of wound size plays a vital role for clinical wound treatment as the physical dimension of a wound is an important clue for wound assessment. The current techniques for wound area measurement are the ruler method and tracing which is mainly based on visual inspection, thus are not very accurate as well as time-consuming. A computerized wound measurement system can provide a more accurate measurement, reduce bias and errors due to fatigue and can potentially reduce clinical workload. In this paper, we proposed a simple but efficient method for wound area segmentation based on superpixel classification with color and texture feature and SVM classifier. Som...
Research Interests:
Alan Tan Alireza Ahrary AmbarDutta Amjad A.Altadmri Andrews Samraj Arman Sargolzaei Baiza Achmad BeeYanHiew BoZhou Carl James Debono Chilukuri MV Chockalingam Aravind Vaithilingam Choun-Sian Lim Dr Agileswari Ramasamy Dwarikanath... more
Alan Tan Alireza Ahrary AmbarDutta Amjad A.Altadmri Andrews Samraj Arman Sargolzaei Baiza Achmad BeeYanHiew BoZhou Carl James Debono Chilukuri MV Chockalingam Aravind Vaithilingam Choun-Sian Lim Dr Agileswari Ramasamy Dwarikanath Mahapatra Farah Hani Nordin Farah Yasmin Abdul Rahman Fatimah Khalid Fazly Salleh Abas Gueesang Lee Hariharan Muthusamy Hasliza Abu Hassan herlina abdul rahim Hezerul Abdul Karim Hua Nong Ting Ibrahima Faye Ihsan Yassin jingying chen Kalaivani Chellappan Kamel Abderrahim Kar Seng Loke ...
Monitoring the changes in wound size over time is one of the most important parameters for assessment of the applied treatment plan in wound care. The current manual methods for wound measurement have several drawbacks including high... more
Monitoring the changes in wound size over time is one of the most important parameters for assessment of the applied treatment plan in wound care. The current manual methods for wound measurement have several drawbacks including high error rates, time consuming and also causing patient discomfort because of pain or infection. There are some computerized tools available in clinical settings known as digital planimetry which require the clinician to identify wound boundaries and calibrate them manually. However these methods are a barrier to achieving clinical quality benchmarks as the accuracy decreases drastically with improper camera lens orientation and some other factors. This paper proposes an automated calibration method for wound measurement based on the segmentation of reference label card and geometrical image operations. Due to the uncontrolled setting of the camera for our images, segmentation of the label card can be very challenging, thus preventing accurate calibration and wound size measurement. We proposed a label card segmentation method based on probability map approach and superpixel region growing. Theoretically, the white label card regions would have a distinguished probability map in the white channel compared to other regions, generating a reliable initial segmentation. Morphological operations and edge detection are then performed to obtain a more accurate final segmentation. Based on the segmented label card, the size of the pixel can be calibrated, and the size of the wound can be automatically calculated. Experimental results confirm the feasibility of the proposed method by demonstrating segmentation accuracy in terms of the Jaccard index of 94.77%, Dice coefficient of 97.20%, and contour matching score of 94.21%. The wound measurement accuracy also shows very promising results.
The presence of robots in our daily life is becoming more common, where robots start carrying out more complex tasks. This increase in the complexity of tasks makes conventional control system insufficient. Therefore, a plausible approach... more
The presence of robots in our daily life is becoming more common, where robots start carrying out more complex tasks. This increase in the complexity of tasks makes conventional control system insufficient. Therefore, a plausible approach is required for robots to learn how to perform these tasks. Reinforcement learning enables robots to perform complex tasks without highly engineered control systems. However, using reinforcement learning in robotic applications is challenged by several problems such as high dimensionality. Thus, in this paper, we study the performance of the Hindsight Experience Replay (HER) algorithm which addresses the high dimensionality problem. In this paper, we analyze the algorithm performance using a simulated robotic arm to pick and place different objects. Then, we propose the use of vision feedback which is used to control the gripper of the robotic arm. The results and analysis highlights some of HER limitations when dealing with objects that have limited grasping points. Our proposed method allows the robotic arm to pick objects using the same trained policy without the need to retrain the agent for new objects. Finally, we prove that using our method the robotic arm can pick the objects with higher success rate compared to the one without vision feedback.
An individual with profound deafness or total hearing loss has a hearing threshold of 80dB or more. The ineffectiveness of hearing aids, surging costs and complex surgeries for cochlear implants have discouraged many to opt for these... more
An individual with profound deafness or total hearing loss has a hearing threshold of 80dB or more. The ineffectiveness of hearing aids, surging costs and complex surgeries for cochlear implants have discouraged many to opt for these types of treatments. Hence, this research aims to provide an alternative hearing aid that stimulates “hearing” through the skin sensory, which is more affordable and accessible for the profoundly deaf or total hearing loss community. We have developed four initial vibrating transducers with single spectrum, which are strapped to a belt. The transducers pick up audible sounds through a microphone, amplifies the sound to a high-level signal, stimulating a vibration pattern on the human skin sensory. The belt was tested on 30 random people who identified as normal, partial, and profoundly deaf. When the belt was strapped to the individual’s waist, audible sound was played (stimulus) and the individual was asked whether he/she can feel a stimulation or vi...
Research Interests:
A tremendous increase in the video content uploaded on the internet has made it necessary for auto-recognition of videos in order to analyze, moderate or categorize certain content that can be accessed easily later on. Video analysis... more
A tremendous increase in the video content uploaded on the internet has made it necessary for auto-recognition of videos in order to analyze, moderate or categorize certain content that can be accessed easily later on. Video analysis requires the study of proficient methodologies at the semantic level in order to address the issues such as occlusions, changes in illumination, noise, etc. This paper is aimed at the analysis of the soccer videos and semantic processing as an application in the video semantic analysis field. This study proposes a framework for automatically generating and annotating the highlights from a soccer video. The proposed framework identifies the interesting clips containing possible scenes of interest, such as goals, penalty kicks, etc. by parsing and processing the audio/video components. The framework analyzes, separates and annotates the individual scenes inside the video clips and saves using kernel support vector machine. The results show that semantic a...
Research Interests:
Hormone receptor status in breast carcinoma is determined primarily to identify patients who may benefit from hormonal therapy. Estrogen receptor (ER) is one of the hormone receptor positive factors which have been recognized as a marker... more
Hormone receptor status in breast carcinoma is determined primarily to identify patients who may benefit from hormonal therapy. Estrogen receptor (ER) is one of the hormone receptor positive factors which have been recognized as a marker for which women with breast cancer would respond to hormone treatment. We propose a system to classify cells in ER-stained whole slide breast carcinoma images according to their staining strength using convolutional neural network (CNN). The proposed CNN multiclass classifier was tested on a region of 1200 cells, and achieved very promising results, with overall accuracy of 88.8% and AUC score of 97.5%. The proposed system is useful for use in hormone receptor testing, where the outcomes are used to decide whether the cancer is likely to respond to hormonal therapy or other treatments.
Research Interests:
Automatic and accurate detection of positive and negative nuclei from images of immunostained tissue biopsies is critical to the success of digital pathology. The evaluation of most nuclei detection algorithms relies on manually generated... more
Automatic and accurate detection of positive and negative nuclei from images of immunostained tissue biopsies is critical to the success of digital pathology. The evaluation of most nuclei detection algorithms relies on manually generated ground truth prepared by pathologists, which is unfortunately time-consuming and suffers from inter-pathologist variability. In this work, we developed a digital immunohistochemistry (IHC) phantom that can be used for evaluating computer algorithms for enumeration of IHC positive cells. Our phantom development consists of two main steps, 1) extraction of the individual as well as nuclei clumps of both positive and negative nuclei from real WSI images, and 2) systematic placement of the extracted nuclei clumps on an image canvas. The resulting images are visually similar to the original tissue images. We created a set of 42 images with different concentrations of positive and negative nuclei. These images were evaluated by four board certified patho...
Research Interests:
HSP90, a highly conserved molecular chaperone that regulates the function of several oncogenic client proteins is altered in glioblastoma. However, HSP90 inhibitors currently in clinical trials are short-acting, have unacceptable... more
HSP90, a highly conserved molecular chaperone that regulates the function of several oncogenic client proteins is altered in glioblastoma. However, HSP90 inhibitors currently in clinical trials are short-acting, have unacceptable toxicities or are unable to cross the blood brain barrier (BBB). We examined the efficacy of onalespib, a potent, long-acting novel HSP90 inhibitor as a single agent and in combination with temozolomide (TMZ) against gliomas invitro and invivo<br /><br />Experimental Design: The effect of onalespib on HSP90, its client proteins, and on the biology of glioma cell lines and patient-derived glioma-initiating cells (GSC) was determined. Brain and plasma pharmacokinetics of onalespib and its ability to inhibit HSP90 in vivo was assessed in non-tumor bearing mice. Its efficacy as a single agent or in combination with TMZ was assessed in vitro and in vivo using zebrafish and patient derived GSC xenograft mouse glioma models.<br /><br />Resu...
Research Interests:
The advance of high resolution digital scans of pathology slides allowed development of computer based image analysis algorithms that may help pathologists in IHC stains quantification. While very promising, these methods require further... more
The advance of high resolution digital scans of pathology slides allowed development of computer based image analysis algorithms that may help pathologists in IHC stains quantification. While very promising, these methods require further refinement before they are implemented in routine clinical setting. Particularly critical is to evaluate algorithm performance in a setting similar to current clinical practice. In this article, we present a pilot study that evaluates the use of a computerized cell quantification method in the clinical estimation of CD3 positive (CD3+) T cells in follicular lymphoma (FL). Our goal is to demonstrate the degree to which computerized quantification is comparable to the practice of estimation by a panel of expert pathologists. The computerized quantification method uses entropy based histogram thresholding to separate brown (CD3+) and blue (CD3-) regions after a color space transformation. A panel of four board-certified hematopathologists evaluated a d...
Research Interests:
Acne is a common skin condition present predominantly in the adolescent population, but may continue into adulthood. Scarring occurs commonly as a sequel to severe inflammatory acne. The presence of acne and resultant scars are more than... more
Acne is a common skin condition present predominantly in the adolescent population, but may continue into adulthood. Scarring occurs commonly as a sequel to severe inflammatory acne. The presence of acne and resultant scars are more than cosmetic, with a significant potential to alter quality of life and even job prospects. The psychosocial effects of acne and scars can be disturbing and may be a risk factor for serious psychological concerns. Treatment efficacy is generally determined based on an invalidated gestalt by the physician and patient. However, the validated assessment of acne can be challenging and time consuming. Acne can be classified into several morphologies including closed comedones (whiteheads), open comedones (blackheads), papules, pustules, cysts (nodules) and scars. For a validated assessment, the different morphologies need to be counted independently, a method that is far too time consuming considering the limited time available for a consultation. However, it is practical to record and analyze images since dermatologists can validate the severity of acne within seconds after uploading an image. This paper covers the processes of region-ofinterest determination using entropy-based filtering and thresholding as well acne lesion feature extraction. Feature extraction methods using discrete wavelet frames and gray-level co-occurence matrix were presented and their effectiveness in separating the six major acne lesion classes were discussed. Several classifiers were used to test the extracted features. Correct classification accuracy as high as 85.5% was achieved using the binary classification tree with fourteen principle components used as descriptors. Further studies are underway to further improve the algorithm performance and validate it on a larger database.
Research Interests:
Intraoperative neuropathology of glioma recurrence represents significant visual challenges to pathologists as they carry significant clinical implications. For example, rendering a diagnosis of recurrent glioma can help the surgeon... more
Intraoperative neuropathology of glioma recurrence represents significant visual challenges to pathologists as they carry significant clinical implications. For example, rendering a diagnosis of recurrent glioma can help the surgeon decide to perform more aggressive resection if surgically appropriate. In addition, the success of recent clinical trials for intraoperative administration of therapies, such as inoculation with oncolytic viruses, may suggest that refinement of the intraoperative diagnosis during neurosurgery is an emerging need for pathologists. Typically, these diagnoses require rapid/STAT processing lasting only 20-30 minutes after receipt from neurosurgery. In this relatively short time frame, only dyes, such as hematoxylin and eosin (H and E), can be implemented. The visual challenge lies in the fact that these patients have undergone chemotherapy and radiation, both of which induce cytological atypia in astrocytes, and pathologists are unable to implement helpful biomarkers in their diagnoses. Therefore, there is a need to help pathologists differentiate between astrocytes that are cytologically atypical due to treatment versus infiltrating, recurrent, neoplastic astrocytes. This study focuses on classification of neoplastic versus non-neoplastic astrocytes with the long term goal of providing a better neuropathological computer-aided consultation via classification of cells into reactive gliosis versus recurrent glioma. We present a method to detect cells in H and E stained digitized slides of intraoperative cytologic preparations. The method uses a combination of the ‘value’ component of the HSV color space and ‘b*’ component of the CIE L*a*b* color space to create an enhanced image that suppresses the background while revealing cells on an image. A composite image is formed based on the morphological closing of the hue-luminance combined image. Geometrical and textural features extracted from Discrete Wavelet Frames and combined to classify cells into neoplastic and non-neoplastic categories. Experimental results show that there is a strong consensus between the proposed method’s cell detection markings with those of the pathologist’s. Experiments on 48 images from six patients resulted in F1-score as high as 87.48%, 88.08% and 86.12% for Reader 1, Reader 2 and the reader consensus, respectively. Classification results showed that for both readers, binary classification tree and support vector machine performed the best with F1-scores ranging 0.92 to 0.94.
Research Interests:
Locally Linear Embedding (LLE) is an unsupervised non-linear manifold learning method, which has spurred increased interest in face recognition research recently. However, it is commonly known that a supervised method that considering the... more
Locally Linear Embedding (LLE) is an unsupervised non-linear manifold learning method, which has spurred increased interest in face recognition research recently. However, it is commonly known that a supervised method that considering the class-specific information always outperforms the unsupervised one, especially in biometric recognition task. In this paper, we propose a supervised LLE technique, known as class-label Locally Linear Embedding (cLLE). cLLE aims to discover the nonlinearity of high-dimensional data by minimizing the global reconstruction error of the set of all local neighbours in the data set. cLLE method is using user class-specific information in neighbourhoods selection and thus preserves the local neighbourhoods. Since the locality preservation is correlated to the class discrimination, the proposed cLLE is expected superior to LLE in face recognition. Experimental results on three face databases demonstrate the success of the proposed technique.
This paper presents a method to recognize badminton action from depth map sequences acquired by Microsoft Kinect sensor. Badminton is one of Malaysia’s most popular, but there is still lack of research on action recognition focusing on... more
This paper presents a method to recognize badminton action from depth map sequences acquired by Microsoft Kinect sensor. Badminton is one of Malaysia’s most popular, but there is still lack of research on action recognition focusing on this sport. In this research, bone orientation details of badminton players are computed and extracted in order to form a bag of quaternions feature vectors. After conversion to log-covariance matrix, the system is trained and the badminton actions are classified by a support vector machine classifier. Our experimental dataset of depth map sequences composed of 300 badminton action samples of 10 badminton actions performed by six badminton players. The dataset varies in terms of human body size, clothes, speed, and gender. Experimental result has shown that nearly 92% of average recognition accuracy (ARA) was achieved in inter-class leave one sample out cross validation test. At the same time, 86% of ARA was achieved in inter-class cross subject valid...
Research Interests:
Two C-band RADARSAT synthetic aperture radar (SAR) images are used to automatically detect changes that occured in between year 2004 and 2008 due to extensive deforestation, at Gadong River area of Belum-Temengor, Northern region of West... more
Two C-band RADARSAT synthetic aperture radar (SAR) images are used to automatically detect changes that occured in between year 2004 and 2008 due to extensive deforestation, at Gadong River area of Belum-Temengor, Northern region of West Malaysia. Changes are detected by automatically finding the best threshold value of the standard log-ratio image that is derived from the two multitemporal SAR
Research Interests:
Gestures are indeed important in our daily life as they serve as one of the communication platform by using body motions in order to deliver information or effectively interact. This paper proposes to leverage the Kinect sensor for... more
Gestures are indeed important in our daily life as they serve as one of the communication platform by using body motions in order to deliver information or effectively interact. This paper proposes to leverage the Kinect sensor for close-range human gesture recognition. The orientation details of human arms are extracted from the skeleton map sequences in order to form a bag of quaternions feature vectors. After the conversion to log-covariance matrix, the system is trained and the gestures are classified by multi-class SVM classifier. An experimental dataset of skeleton map sequences for 5 subjects with 6 gestures was collected and tested. The proposed system obtained remarkably accurate result with nearly 99 % of average correct classification rate (ACCR) compared to state of the art method with ACCR of 95 %.
Visual tracking is still an open problem because one needs to discriminate between the target object and background under long duration. There is a major problem with conventional adaptive tracking where the target object is incorrectly... more
Visual tracking is still an open problem because one needs to discriminate between the target object and background under long duration. There is a major problem with conventional adaptive tracking where the target object is incorrectly learnt (adapted) during runtime, resulting in poor performance of tracker. In this paper, we address this problem by proposing validation-update strategy to minimize the error of false patches updating. The classifier we use is based on boosted ensemble of Local Dominant Orientation (LDO). However, since LDO features contain binary values which are unsuitable for classification, we have added a process to the online boosting learning algorithm that permits the two binary values of "0" and "1". We elevate the tracker performance by pairing the classifier with normalized crosscorrelation of patches tracked by Lukas-Kanade tracker. In the experiment conducted, we compare our method with two other state-of-the-art adaptive trackers using BoBot dataset. Our method yields good tracking performance under variety of scenarios set by BoBot dataset.
In sports, players have to perform body movement in a specific manner in order to obtain desired training effect. Badminton is one of Malaysia’s most popular sport, but there is still lack of research on action analysis focusing on this... more
In sports, players have to perform body movement in a specific manner in order to obtain desired training effect. Badminton is one of Malaysia’s most popular sport, but there is still lack of research on action analysis focusing on this sport. In this paper, a method to analyze badminton action from depth map sequences acquired by Microsoft Kinect sensor is proposed. A compact and view invariant representation of the joint movement, namely region of movement index is generated from the three-dimensional coordinates of the tracked joint. Then, the overall or frame-based performance variation between the expert and learner is computed by mapping using dynamic time warping algorithm.
Research Interests:
An advance in today's information communication technology (ICT) has opened an opportunity for healthcare industry to enhance and improve the quality of their services. Decision makers are now able to make their decisions accurately... more
An advance in today's information communication technology (ICT) has opened an opportunity for healthcare industry to enhance and improve the quality of their services. Decision makers are now able to make their decisions accurately and precisely with the helps of computer-based decision support system. Sharing of patient's medical records is not a new approach. From paper based format to electronic format, patient's medical records are shared among physicians, medical staffs and etc. Shared data might be carried manually from one department to another department or sent physically from hospital to hospital. With Internet technology, some medical data were sent via e-mail. These methods encountered many problems to patients and practitioners. Security and privacy issues, data lost, and delivery durations were some of the concerns that should be overcome. This paper will discussed an idea on how to overcome above mentioned issues and proposed a solution that can be served as the platform for future medical data sharing in telemedicine. The successful development of the working prototype will greatly enhance the functionality of existing data sharing in the hospital. At the same time, the tools and algorithms designed in this idea will helps to solve some of the data mining challenges
Research Interests:
Research Interests:
Abstract Locally Linear Embedding (LLE), which has recently emerged as a powerful face feature descriptor, suffers from a limitation. That is class-specific information of data is lacked of during face analysis. Thus, we propose a... more
Abstract Locally Linear Embedding (LLE), which has recently emerged as a powerful face feature descriptor, suffers from a limitation. That is class-specific information of data is lacked of during face analysis. Thus, we propose a supervised LLE technique, known as class-label Locally Linear Embedding (cLLE), to overcome the problem. cLLE is able to discover the nonlinearity of high-dimensional face data by minimizing the global reconstruction error of the set of all local neighbors in the data set. cLLE utilizes user class-specific information in ...
Research Interests:
Research Interests:
For face recognition, graph embedding techniques attempt to produce a high data locality projection for better recognition performance. However, estimation of population data locality could be severely biased due to small number of... more
For face recognition, graph embedding techniques attempt to produce a high data locality projection for better recognition performance. However, estimation of population data locality could be severely biased due to small number of training samples. The biased estimation triggers overfitting problem and hence poor generalization. In this paper, we propose a new linear graph embedding technique based upon an adaptive locality preserving regulation model (ALPRM), known as Regularized Locality Preserving Discriminant Embedding ( ...
Research Interests:
This paper focuses on two cardiac conditions, the supraventricular ectopy and the ventricular ectopy. Four different mother wavelets are used to produce sets of features. Results shows that each cardiac conditions beat has its own unique... more
This paper focuses on two cardiac conditions, the supraventricular ectopy and the ventricular ectopy. Four different mother wavelets are used to produce sets of features. Results shows that each cardiac conditions beat has its own unique characteristics and also decomposition of different mother wavelet produced different degree in discriminative power. The Discriminant Analysis Classifier of different distance metric (linear, quadratic
Research Interests:
Research Interests:
Hahn moments are a superset of Tchebichef and Krawtchouk moments. The formulation for Hahn moments is however comparably more complex than other moments. So far only research work on translation and scale invariants for Tchebichef moments... more
Hahn moments are a superset of Tchebichef and Krawtchouk moments. The formulation for Hahn moments is however comparably more complex than other moments. So far only research work on translation and scale invariants for Tchebichef moments has been presented but not on Hahn moments. In this paper, a moment normalization method to achieve translation and scale invariants of Hahn moments is proposed. This method applies the concept of mapping functions used in image normalization. The mapping functions, once determined, are plugged into the moment generating functions to generate moment invariants. The proposed method is simpler and flexible. Experimental results show that faster execution and more precise moment invariants can be achieved using the invariant generating functions.
Research Interests:
Research Interests:
Research Interests:
Research Interests: Cognitive Science, Archaeology, Algorithms, Computer Graphics, Image Processing, and 15 moreArt, Culture, Archives, Image Analysis, Hypermedia, Database Management Systems, Content based image retrieval, Image Retrieval, Computer User Interface Design, Information Storage and Retrieval, Information Dissemination, Image Enhancement, Reproducibility of Results, Internet, and Electrical And Electronic Engineering
ABSTRACT A new approach to image retrieval is presented in the domain of museum and gallery image collections. Specialist algorithms, developed to address specific retrieval tasks, are combined with more conventional content and metadata... more
ABSTRACT A new approach to image retrieval is presented in the domain of museum and gallery image collections. Specialist algorithms, developed to address specific retrieval tasks, are combined with more conventional content and metadata retrieval approaches, and implemented within a distributed architecture to provide cross-collection searching and navigation in a seamless way. External systems can access the different collections using interoperability protocols and open standards, which were extended to accommodate content based as well as text based retrieval paradigms. After a brief overview of the complete system, we describe the novel design and evaluation of some of the specialist image analysis algorithms including a method for image retrieval based on sub-image queries, retrievals based on very low quality images and retrieval using canvas crack patterns. We show how effective retrieval results can be achieved by real end-users consisting of major museums and galleries, accessing the distributed but integrated digital collections.
Research Interests:
... If both parties are aware, it will create a win-win situation which indirectly creates the K-Sharing culture. ... UM 2007 T3-21 CHALLENGES OF MEASURING USER SATISFACTION VIA AUTOMATIC FACIAL EXPRESSION ANALYSIS Zolidah Kasiran and... more
... If both parties are aware, it will create a win-win situation which indirectly creates the K-Sharing culture. ... UM 2007 T3-21 CHALLENGES OF MEASURING USER SATISFACTION VIA AUTOMATIC FACIAL EXPRESSION ANALYSIS Zolidah Kasiran and Saadiah Yahya Faculty of ...
Two C-band RADARSAT synthetic aperture radar (SAR) images are used to automatically detect changes that occured in between year 2004 and 2008 due to extensive deforestation, at Gadong River area of Belum-Temengor, Northern region of West... more
Two C-band RADARSAT synthetic aperture radar (SAR) images are used to automatically detect changes that occured in between year 2004 and 2008 due to extensive deforestation, at Gadong River area of Belum-Temengor, Northern region of West Malaysia. Changes are detected by automatically finding the best threshold value of the standard log-ratio image that is derived from the two multitemporal SAR
Research Interests:
Visual tracking is still an open problem because one needs to discriminate between the target object and background under long duration. There is a major problem with conventional adaptive tracking where the target object is incorrectly... more
Visual tracking is still an open problem because one needs to discriminate between the target object and background under long duration. There is a major problem with conventional adaptive tracking where the target object is incorrectly learnt (adapted) during runtime, resulting in poor performance of tracker. In this paper, we address this problem by proposing validation-update strategy to minimize the error of false patches updating. The classifier we use is based on boosted ensemble of Local Dominant Orientation (LDO). However, since LDO features contain binary values which are unsuitable for classification, we have added a process to the online boosting learning algorithm that permits the two binary values of "0" and "1". We elevate the tracker performance by pairing the classifier with normalized crosscorrelation of patches tracked by Lukas-Kanade tracker. In the experiment conducted, we compare our method with two other state-of-the-art adaptive trackers using BoBot dataset. Our method yields good tracking performance under variety of scenarios set by BoBot dataset.