Todor D Ganchev
Tehnical university varna, Electronics, Faculty Member
- Speech Processing, Speaker Recognition, Feature Extraction, MFCC features, Research Methodology, Bioacoustics, and 9 moreDigital Signal Processing, Pattern Recognition, Machine Learning, Audio Signal Processing, Biodiversity Informatics, Artificial Intelligence, Data Mining, Support vector machine, and Decision Treeedit
- -edit
We present the results from the first large-scale study on pollution in the Varna Lake sediments, which was implemented in two preselected zones with significant importance for developing economic activities and urban planning. Based on... more
We present the results from the first large-scale study on pollution in the Varna Lake sediments, which was implemented in two preselected zones with significant importance for developing economic activities and urban planning. Based on 140 probes, we developed a digital map of the pollution gradients in these zones. The sediment analysis shows that the pollution levels at specific locations significantly exceed the international standards for contamination with heavy metals and petroleum products. The reported pollution concentrations can be linked to various contamination sources, such as rivers, ports, the chemical industry, and WWTPs (Waste water treatment plants), located on the shores of the lake. The high values of pollution with iron, copper, and petroleum product registered in the sediments of Lake Varna indicate the need for sustainable management of human activities in the area. The presented findings motivate urgent measures for cleaning and improving the ecological situ...
Research Interests:
We present a gathering of information what is heart rate variability, it’s uses and interpretation. We also make an overview of the methods in time and frequency domain and also the nonlinear method for calculation of heart rate... more
We present a gathering of information what is heart rate variability, it’s uses and interpretation. We also make an overview of the methods in time and frequency domain and also the nonlinear method for calculation of heart rate variability. The experiments compare different implementations of those methods in separate platforms – Matlab and the widely used, among medical experts and specialist, software platform Kubios. The results in the time domain prove to be almost identical, the frequency domain show similar trends even if the absolute values are not identical, and the results in the nonlinear calculations are similar.
Research Interests:
As the population ageing is one of the most significant trends of the 21st century, the traditional methods for monitoring and diagnosis of elderly people are no longer sufficient for providing them with a satisfactory quality of life.... more
As the population ageing is one of the most significant trends of the 21st century, the traditional methods for monitoring and diagnosis of elderly people are no longer sufficient for providing them with a satisfactory quality of life. One of the promising approaches that helps towards addressing such challenges is the employment of communication and information-based systems, allowing automated remote health monitoring and data analysis. The evolutionary transition to digital healthcare requires the definition of new concepts, architectural models and functionalities, subject of study by the scientific community and technology standardization bodies. In the present paper, we overview a wide range of architectures and technological solutions that were recently proposed for the needs of health monitoring, with focus on systems that are based on communication connectivity. We provide a comprehensive analysis and some critical comments on the potential technological shortcomings and challenges of these solutions.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
In the present work, we consider the automated detection of negative emotions and high-arousal negativevalence (HANV) states, which are akin to acute stress occurring in specific context. We investigate the influence and intricacy which... more
In the present work, we consider the automated detection of negative emotions and high-arousal negativevalence (HANV) states, which are akin to acute stress occurring in specific context. We investigate the influence and intricacy which subjective perception of negative emotions and acute stress brings to the process of automated recognition of these states. For that purpose we experimentally evaluate the advantages of modelling based on the person-independent tags of the audio-video stimuli, and models built with the personspecific self-reported tags. Based on the self-reported tags, which are obtained with only tiny extra effort, we report a relative improvement of the HANV detection accuracy with up to 5%.
Research Interests:
The Locally Recurrent Probabilistic Neural Network (LRPNN) consists of an input layer, three hidden layers and an output layer. The first two hidden layers are derived from the original PNN, while the third layer referred as recurrent... more
The Locally Recurrent Probabilistic Neural Network (LRPNN) consists of an input layer, three hidden layers and an output layer. The first two hidden layers are derived from the original PNN, while the third layer referred as recurrent layer is capable to model correlations within temporal sequences of observations. In the present study, we investigate the feasibility of FPGA-based implementation of the locally recurrent layer of LRPNN. An important consideration due to the specifics of this architecture is the use of modules with very high precision in the hardware design. Although expensive in terms of available resources in the FPGA chip, this is necessary, in order to compensate for the added error of quantization due to the multiple feedbacks from neurons in the neural network. The weights for the recurrent layer of the LRPNN are automatically computed from the available training data and translated into the hardware design. The experimental evaluation was carried out on the DEAP database, where two classes of emotional states were considered. The design makes use of a computed short-term energy from a 32-channel electroencephalographic (EEG) signal as an input. Results of an extensive experimental validation show that there is approximately one percent difference between the accuracy achieved with CPU-based software and FPGA-based hardware implementation of the LRPNN.
Research Interests:
We present the overall design and implementation of a wearable system that is capable to continuously monitor and register negative emotions, high levels of emotion arousal, and high-arousal-negative-valence states from physiological... more
We present the overall design and implementation of a wearable system that is capable to continuously monitor and register negative emotions, high levels of emotion arousal, and high-arousal-negative-valence states from physiological signals, such as skin conductivity and ECG. This system builds on the client-server architecture. The commercially available wireless data acquisition devices Shimmer$3~\mathrm {G}\mathrm {S}\mathrm {R}+$ and Shimmer3 ECG are used for the acquisition of physiological signals, which are then transmitted over a Bluetooth channel to a mobile phone. The mobile phone hosts the user interface and implements the data aggregation and transmission to the server, which carries all signal processing and classification tasks. Purposely developed software implements all data processing and recognition tasks on the server side and the user interface and the data communication on the client side. We report evaluation results for various setups of the binary detectors of negative emotions, high level of emotional arousal, and high-arousal- negative-valence states.
Research Interests:
We present a computationally efficient non-parametric algorithm for the automated detection of systolic peaks in photoplethysmography (PPG) signal that does not require preprocessing for artifact elimination, signal filtering, or... more
We present a computationally efficient non-parametric algorithm for the automated detection of systolic peaks in photoplethysmography (PPG) signal that does not require preprocessing for artifact elimination, signal filtering, or detrending. It is validated in an experimental setup based on the publicly available CLAS dataset. The experimental results show that it outperforms two well-known methods in terms of detection accuracy and computational demands. We report a very high detection accuracy, with an error rate below 0.5%, on good quality signals and below 13% on very low-quality PPG signals. The proposed algorithm is characterized with very short processing times and on a low-cost laptop computer requires approximately 0.000012 real-time for the processing of a 60-seconds recording.
Research Interests:
Research Interests:
We present the conceptual design of an e-Learning framework that makes use of an advanced technology in support of the personalized training of young people. This concept builds on the integration of novel functionality that implements an... more
We present the conceptual design of an e-Learning framework that makes use of an advanced technology in support of the personalized training of young people. This concept builds on the integration of novel functionality that implements an automated objective evaluation of the cognitive effort, the degree of attention and concentration, the emotional condition, and the stress level of a trainee, based on physiological signal processing. Such an awareness helps for the automated adaptation of the training process to the personality traits and to the momentary capacity of each trainee. Such a novel technology allows the development of intelligent training applications that adapt to the momentary capacity of the trainee to comprehend and advance with the current tasks and help towards the optimization of the learning outcome. Such a concept also offers new opportunities for the development of more engaging and more efficient training process and the design of an universally accessible i...
Research Interests:
Research Interests:
We present the design of a mobile system for real-time stress-level assessment. The system combines wearable sensors, wireless data acquisition, and Cloud computing in order to collect and analyze physiological signals, such as, Galvanic... more
We present the design of a mobile system for real-time stress-level assessment. The system combines wearable sensors, wireless data acquisition, and Cloud computing in order to collect and analyze physiological signals, such as, Galvanic Skin Response (GSR) and skin temperature. We report on the implementation of a specific use case, which incorporates functionality for real-time data logging and analysis. Experimental results demonstrate excellent recognition accuracy of affective arousal and decent accuracy for binary detection of valence. In addition, we also evaluate the feasibility for detection of high arousal/negative valence (HANV) events, which in specific setups can be connected to stress.
Research Interests:
In the presents paper, we summarize the results of a recent study on the relationship between task-caused acute stress and students' performance. Our experimental protocol was based on the CLAS dataset, which contains physiological... more
In the presents paper, we summarize the results of a recent study on the relationship between task-caused acute stress and students' performance. Our experimental protocol was based on the CLAS dataset, which contains physiological signals of 60 students. The physiological recordings were captured during students' involvement in five different tasks, including three interactive tasks (Stroop test, Math test, IQ test) and two noninteractive tasks. The non-interactive tasks aimed at emotion elicitation via blocks of sixteen photographs and sixteen emotional music video clips, purposely selected to cover the entire arousal-valence space. We observed that the three Stroop, Math and IQ tests cause higher acute stress levels than pictures and musical stimuli purposely selected to provoke emotional reactions. The experimental results show that acute stress has different effects on students' academic performance, depending on their gender and individually. Specifically, in contrast to the females, males were observed to show lower stress levels in the Math test. At the same time, males were observed to be less concentrated on the Stoop test. We observed that heart-rate variability (HRV) could be used as an indicator of the students' performance under stress as it is not related to students' abilities.
Research Interests:
We present the overall design and the implementation of the CLAS dataset, a multimodal resource which was purposely developed in support of research and technology development (RTD) activities oriented towards the automated recognition of... more
We present the overall design and the implementation of the CLAS dataset, a multimodal resource which was purposely developed in support of research and technology development (RTD) activities oriented towards the automated recognition of some specific states of mind. Although the particular focus of our research is on the states of mind associated with negative emotions, mental strain and high cognitive effort, the CLAS dataset could offer an adequate support to research of a wider scope, such as general studies on attention assessment, cognitive load assessment, emotion recognition, as well as stress detection. The dataset consists of synchronized recordings of physiological signals, such as Electrocardiography (ECG), Plethysmography (PPG), ElectroDermal Activity (EDA), as well as accelerometer data, and metadata of 62 healthy volunteers, which were recorded while involved in three interactive tasks and two perceptive tasks. The interactive tasks aim to elicit different types of cognitive effort and included solving sequences of Math problems, Logic problems and the Stroop test. The perceptive tasks make use of images and audio-video stimuli, purposely selected to evoke emotions in the four quadrants of the arousal-valence space. The joint analysis of success rates in the interactive tasks and the information acquired through the questionnaire and the physiological recordings enables for a multifaceted evaluation of specific states of mind. These results are important for the advancement of research on efficient human-robot collaborations and general research on intelligent human-machine interaction interfaces.
Research Interests:
Research Interests:
In the following paper we present a study on novel signal descriptors for the purposes of automated recognition of negative emotional states from EEG signals - namely, the decorrelated values of the energy of the spatio-temporal... more
In the following paper we present a study on novel signal descriptors for the purposes of automated recognition of negative emotional states from EEG signals - namely, the decorrelated values of the energy of the spatio-temporal distribution of EEG activity. Using the extracted features person-specific SVM models are created. The experimental setups are based on data taken from the DEAP database. The classification accuracy of the proposed features is evaluated using two experimental setups: valence detection and like/dislike detection. Recognition accuracy of 77.5% and 78.0%, respectfully, was achieved.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
In this paper an architecture for the classification of Schizophrenia using EEG-based brain connectivity is proposed. Functional and effective networks were constructed from the EEG using a variety of connectivity measures and with graph... more
In this paper an architecture for the classification of Schizophrenia using EEG-based brain connectivity is proposed. Functional and effective networks were constructed from the EEG using a variety of connectivity measures and with graph theory metrics complex network features were extracted. Several classification algorithms were used for the evaluation of the architecture. Promising results were observed when using connectivity measures that also capture directionality properties of the network. The best classification accuracy was 82.36% and was achieved by Random Forest classifier with Direct Transfer Function as a connectivity measure.
Research Interests:
The study of physiological signals and the evaluation of their features are of great importance for the automated emotion detection, as these are directly connected with the successful modelling and classification of the states of... more
The study of physiological signals and the evaluation of their features are of great importance for the automated emotion detection, as these are directly connected with the successful modelling and classification of the states of interest. In the presented work, we present an evaluation of the appropriateness of LFCC and the logarithmic energy of signals as features for automated recognition of negative emotional states in terms of recognition accuracy. In particular, three sets of features are compared – features computed after frame-level segmentation of the signal; features computed after averaging of frame level descriptors; and features extracted from an entire EEG recording. The performance of the extracted features is evaluated using C4.5 classifier for 10, 15, 20, 30, 45, and 60 filters.
Research Interests:
We present a method for the automated recognition of birdsong syllables type. For that purpose, we automatically segment the birdsong to acoustic evens, and subsequently each segment is modeled by a GMM-based interpolation of the... more
We present a method for the automated recognition of birdsong syllables type. For that purpose, we automatically segment the birdsong to acoustic evens, and subsequently each segment is modeled by a GMM-based interpolation of the short-term energy of the dominant frequency component. Next, the parameters of the GMM model are fed to a classifier in order to recognize the birdsong syllables type. We evaluated the practical worth of this approach using publicly available field recordings of species Myrmotherula multostriata which were recorded in natural habitats. The experimental protocol was based on seventy-eight acoustic events, which were obtained after the automatic segmentation of the audio signal. We report recognition accuracy of up to 98%, depending on the classification method and the particular syllable type. Summarizing the experimental results obtained in this study, we concluded that the proposed method has good potential for achieving higher recognition accuracy, however additional work is needed for satisfying the needs of practical applications.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Varna Lake is of great social and economic importance for the city of Varna and the entire region. The urbanisation of the areas surrounding the Varna-Beloslav Lake complex and the accelerated industrialisation around the coastline of... more
Varna Lake is of great social and economic importance for the city of Varna and the entire region. The urbanisation of the areas surrounding the Varna-Beloslav Lake complex and the accelerated industrialisation around the coastline of Varna Bay have significant anthropogenic pressure on the aquatic ecosystem. An essential aspect is that industrial development, maritime transport and urbanisation are substantial sources of large-scale inorganic and organic pollution with all the resulting negative consequences. Recent measurements show that the surface sediments, which are a source of nutrients for the water body of Varna Lake, are contaminated with heavy metals and oil products. In this paper, we present a quantitative assessment of the degree of contamination with heavy metals (Hg, As, Fe, Cu, Pb) and petroleum products in the upper sediments of Varna Lake and then discuss opportunities for using some of these pollutants as a resource. The exploitation of these resources would brin...
Research Interests:
Research Interests:
In the presented paper we investigate the properties of spectral EEG features for the detection of negative emotional states. In particular, the proposed features represent the dynamics of energy distribution in the frequency range of... more
In the presented paper we investigate the properties of spectral EEG features for the detection of negative emotional states. In particular, the proposed features represent the dynamics of energy distribution in the frequency range of 20–35 Hz, based on a time-frequency analysis of multichannel EEG signal. The experimental evaluation is based on data from the DEAP database. We report results with J48- and SMO-based classifiers, in terms of average classification accuracy, 94.3% and 96.8%, respectively.
Research Interests:
We present a three-step method for attribute selection that builds on person-independent and person-specific feature assessment stages. The first two steps aim to select a person-independent subset of attributes that are repeatedly... more
We present a three-step method for attribute selection that builds on person-independent and person-specific feature assessment stages. The first two steps aim to select a person-independent subset of attributes that are repeatedly selected for a large population of users. Next, this selection is intersect with a person-specific subset derived from the Fisher's separation criterion. As a result, we obtain a subset of attributes which is both task-specific and customized to the quality of data of each particular user. The proposed method was validated on the ASCERTAIN database in an experimental setup oriented towards high-arousal negative-valence detection based on physiological signals. The experimental results support that the proposed method offers advantage in terms of detection accuracy when compared to other subset selection strategies.
Research Interests:
In this paper, we present a new concept of a multimodal Affective Tutoring System (mATS) based on knowledge, which is aimed at covering the entire functionalities of a human teacher - namely training and testing, and personal attitude... more
In this paper, we present a new concept of a multimodal Affective Tutoring System (mATS) based on knowledge, which is aimed at covering the entire functionalities of a human teacher - namely training and testing, and personal attitude towards every student. The system incorporates models of the student and of the task, as well as evaluation module for constant determination of the current states of both. Moreover, based on multimodal input signals and system-generated ones (performance, response time, etc.), supplemented with prior knowledge about the student (self-reported parameters, such as age, gender, level of experience, etc.) and user profiling approach mATS gives a variety of options for realization of adequate adaptive strategies.
Research Interests:
In this study, we report on ongoing research aimed at developing an intelligent architecture for real-time health monitoring that adapts to the 3G/4G/5G communication context. It builds on the integration of (i) intelligent sensor devices... more
In this study, we report on ongoing research aimed at developing an intelligent architecture for real-time health monitoring that adapts to the 3G/4G/5G communication context. It builds on the integration of (i) intelligent sensor devices capable of registering specific behavioural indicators, (ii) mobile personal devices, which serve primarily as data concentrators, information gateways and displays, and (iii) remote storage and processing resources, where health data are aggregated for advanced statistical processing. This architecture builds on the principles of distributed computing and resource storing at the network edge, yet it also relies on methods and technologies typically used in cloud-computing applications and 5G technology. More importantly, however, it is capable to sense the communication context and providing a seamless transition between environments with dissimilar communication capacities and allowing for graceful degradation of functionality in case of a shortage of battery or local storage resources. The last is expected to provide numerous benefits that are expected to facilitate the pervasiveness of eHealth monitoring solutions.
Research Interests:
In the present study we evaluate the performance of various training schemes for the locally recurrent probabilistic neural network and seek for advantageous tradeoffs between required training time and classification accuracy.... more
In the present study we evaluate the performance of various training schemes for the locally recurrent probabilistic neural network and seek for advantageous tradeoffs between required training time and classification accuracy. Specifically, we consider training schemes which make use of a simple incremental procedure for adjusting sigma, as well as methods based on particle swarm optimization or differential evolution in different configurations. The experimental evaluation was carried out in common experimental protocol based on the Parkinson speech dataset. The experimental results show that with a proper training configuration a high accuracy can be achieved even with limited training data.
Research Interests:
Stress is widely associated with increased health risks (heart and brain diseases, diabetes, cancer, behavioural disorders, etc.). Moreover, a prolonged exposure to stress is known to negatively affect work performance, attitude,... more
Stress is widely associated with increased health risks (heart and brain diseases, diabetes, cancer, behavioural disorders, etc.). Moreover, a prolonged exposure to stress is known to negatively affect work performance, attitude, decision-making, etc. Furthermore, monitoring of stress levels and proper stress management are of crucial importance for fire-fighters, rescue crews, police force and other high-risk professions in terms of mission success and workforce preservation. In this regard, here we overview the state of the art in personal health monitoring systems and discuss the overall architecture and technology involved in the implementation of such functionality. A particular focus is put on the technology involved in the assessment of brain activity and negative emotional states, which are linked to stress, behavioural, mental disorders, etc. From application point of view we discuss the technological feasibility of stationary and mobile setups for stress-level assessment and monitoring. Finally, we outline the current trends and future research directions and comment on some inherent limitations of stress-level monitoring and on some challenges that remain unaddressed.
Research Interests:
The success of automated distress detection to a large extent depends on the proper choice of machine learning methods and the appropriate representation of data. In the present study, we evaluate five methods for the synthesis of... more
The success of automated distress detection to a large extent depends on the proper choice of machine learning methods and the appropriate representation of data. In the present study, we evaluate five methods for the synthesis of characteristic descriptors that are appropriate in automated distress detection. These allow efficient data representations and contribute towards a significant reduction of the computational demands. Based on these methods, we synthesized nine alternative characteristic descriptors, which were evaluated in a common experimental protocol. The experimental setup relied on a dataset collected from approximately 6000 oncological patients at different stages of therapy. The dataset consists of the binary responses to specific questions in a purposely-designed questionnaire for self-evaluation of the degree of distress. The experimental results show that the characteristic descriptor referred to as KR8 outperforms the others in terms of detection accuracy without a significant increase in the time required for modeling and classification.
Research Interests:
We report on the development of an automated detector of acute stress based on physiological signals. Our detector discriminates between high and low levels of acute stress accumulated by students when performing cognitive tasks on a... more
We report on the development of an automated detector of acute stress based on physiological signals. Our detector discriminates between high and low levels of acute stress accumulated by students when performing cognitive tasks on a computer. The proposed detector builds on well-known physiological signal processing principles combined with the state-of-art support vector machine (SVM) classifier. The novelty aspects here come from the design and implementation of the signal pre-processing and the feature extraction stages, which were purposely designed and fine-tuned for the specific needs of acute stress detection and from applying existing algorithms to a new problem. The proposed acute stress detector was evaluated in person-specific and person-independent experimental setups using the publicly available CLAS dataset. Each setup involved three cognitive tasks with a dissimilar crux of the matter and different complexity. The experimental results indicated a very high detection ...
Research Interests:
We present an adaptive feature selection method that makes use of Fisher’s discriminant ratio (FDR) with flexible threshold which is adjusted in a person-specific manner. The proposed method is shown to improve the detection of... more
We present an adaptive feature selection method that makes use of Fisher’s discriminant ratio (FDR) with flexible threshold which is adjusted in a person-specific manner. The proposed method is shown to improve the detection of high-arousal negative-valence (HANV) conditions, based on two combinations of physiological signals (ECG+GSR and PPG+GSR). We validate the proposed method in an experimental setup aiming at the automated detection of HANV conditions evoked by audio-visual stimuli and picture stimuli. The experimental results support that the proposed method yields to an improvement of the classification accuracy of an SVM-based detector on average with 5.6%±0.6% in comparison with the traditional non-adaptive FDR-based feature selection using threshold 0.3, and with the full set of 39 features.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
ABSTRACT Intercom headsets are mandatory communication apparatus in high noise environments (HNE). The headset selection in HNE, such as combat vehicles, is crucial for achieving the objectives of communication, as it serves the needs for... more
ABSTRACT Intercom headsets are mandatory communication apparatus in high noise environments (HNE). The headset selection in HNE, such as combat vehicles, is crucial for achieving the objectives of communication, as it serves the needs for both noise reduction and voice reproduction. Although military-grade intercom headsets are typically used under extreme environmental conditions, a standard performance evaluation method exists only for the earphone elements. In the present work we propose an integrated method for the assessment of the electroacoustic performance of HNE headsets in conditions of maximum reproduction level and high environmental noise, focusing on the voice communication quality. Objective methods, such as Automatic Speech Recognition (ASR), Perceptual Evaluation of Speech Quality (PESQ) and Speech Transmission Index (STI) are comparatively evaluated and their results are compared to subjective scores using Multiple Stimuli with Hidden Reference and Anchor (MUSHRA) in order to reveal the best fit metrics.
Research Interests:
Reviews and few non-controlled studies showed the effectiveness of several specific designed computer video-games as an additional form of treatment in several areas. However, there is a lack in the literature of specially designed... more
Reviews and few non-controlled studies showed the effectiveness of several specific designed computer video-games as an additional form of treatment in several areas. However, there is a lack in the literature of specially designed serious-games for treating mental disorders. Playmancer (ICT European initiative) aims to develop and assess a serious videogame that may help to treat underlying processes (e.g. lack of self-control strategies) in Eating and Impulse control disorders. Preliminary data will be shown.
Research Interests: Computer Science, Serious Games, Eating Disorders, Video Games, Self Control, and 15 moreVideo Game, Games for Health, Serious Gaming, Health Studies, Medicine, Binge-eating Disorders, Impulsivity, Library and Information Studies, Humans, Pubmed, Mental Disorder, Serious Game, Pathological Gambling, Public health systems and services research, and Impulse Control
FIGURE 3. Male genitalia of Lerneca inalata beripocone subsp. nov. (paratype). A. Ventral; B. Dorsal; C. Lateral views. Abbreviations: ec.arc–ectophallic arc; ec.ap–ectophallic apodeme; end.sc–endophallic sclerite; ps.s–pseudepiphallic... more
FIGURE 3. Male genitalia of Lerneca inalata beripocone subsp. nov. (paratype). A. Ventral; B. Dorsal; C. Lateral views. Abbreviations: ec.arc–ectophallic arc; ec.ap–ectophallic apodeme; end.sc–endophallic sclerite; ps.s–pseudepiphallic sclerite; p.ps.i–proximal pseudepiphallic invagination; sc.A–sclerite A of pseudepiphallic part; sc.B–sclerite B of pseudepiphallic part; sc.C–sclerite C of pseudepiphallic part; ps.p–pseudepiphallic parameres.
ABSTRACT