Papers by Mihai Gavrilescu
We propose a 3-layer architecture for determining the personality type of a subject by only analy... more We propose a 3-layer architecture for determining the personality type of a subject by only analyzing handwriting. The proposed architecture combines Neural Network and Support Vector Machine approaches and it is tested in various configurations for determining which combination offers the best personality type classification results for each mixture of handwriting features. In order to test the system, we created a new training database based on Myers-Briggs Type Indicator (MBTI) questionnaire with the purpose of eliminating the inconsistencies of the experimental results compared to manual analysis. We present the architecture, the experimental results, as well as further improvements that could be brought to the current architecture.
Using Off-Line Handwriting to Predict Blood Pressure Level: A Neural-Network-Based Approach
We propose a novel, non-invasive, neural-network based, three-layered architecture for determinin... more We propose a novel, non-invasive, neural-network based, three-layered architecture for determining blood pressure levels of individuals solely based on their handwriting. We employ four handwriting features (baseline, lowercase letter “f”, connecting strokes, writing pressure) and the result is computed as low, normal or high blood pressure. We create our own database to correlate handwriting with blood pressure levels and we show that it is important to use a predefined text for the handwritten sample used for training the system in order to have high prediction accuracy, while for further tests any random text can be used, keeping the accuracy at similar levels. We obtained over 84% accuracy in intra-subject tests and over 78% accuracy in inter-subject tests. We also show there is a link between several handwriting features and blood pressure level prediction with high accuracy which can be further exploited to improve the accuracy of the proposed approach.
We present a novel non-invasive neural network based three layered system for detecting fatigue b... more We present a novel non-invasive neural network based three layered system for detecting fatigue by analyzing facial expressions evaluated using the Facial Action Coding System. We analyze 16 Action Units pertaining to eye and mouth regions of the face. We define an Action Units map containing Action Unit intensity levels for each frame in the video sequence and we analyze this map in a pattern recognition task via a feed-forward neural network. We show that emotion-induced frontal face recordings offer more information in the training stage, while for testing stage the random dataset can be used with no major impact on accuracy, specificity and sensitivity. We obtain over 88% accuracy in intra-subject tests and over 83% for inter-subject tests and we show that our system surpasses the state-of-the-art in terms of accuracy, specificity, sensitivity and response time.
Sensors, 2019
We present the first study in the literature that has aimed to determine Depression Anxiety Stres... more We present the first study in the literature that has aimed to determine Depression Anxiety Stress Scale (DASS) levels by analyzing facial expressions using Facial Action Coding System (FACS) by means of a unique noninvasive architecture on three layers designed to offer high accuracy and fast convergence: in the first layer, Active Appearance Models (AAM) and a set of multiclass Support Vector Machines (SVM) are used for Action Unit (AU) classification; in the second layer, a matrix is built containing the AUs’ intensity levels; and in the third layer, an optimal feedforward neural network (FFNN) analyzes the matrix from the second layer in a pattern recognition task, predicting the DASS levels. We obtained 87.2% accuracy for depression, 77.9% for anxiety, and 90.2% for stress. The average prediction time was 64 s, and the architecture could be used in real time, allowing health practitioners to evaluate the evolution of DASS levels over time. The architecture could discriminate wi...
Data, 2019
We propose a novel feedforward neural network (FFNN)-based speech emotion recognition system buil... more We propose a novel feedforward neural network (FFNN)-based speech emotion recognition system built on three layers: A base layer where a set of speech features are evaluated and classified; a middle layer where a speech matrix is built based on the classification scores computed in the base layer; a top layer where an FFNN- and a rule-based classifier are used to analyze the speech matrix and output the predicted emotion. The system offers 80.75% accuracy for predicting the six basic emotions and surpasses other state-of-the-art methods when tested on emotion-stimulated utterances. The method is robust and the fastest in the literature, computing a stable prediction in less than 78 s and proving attractive for replacing questionnaire-based methods and for real-time use. A set of correlations between several speech features (intensity contour, speech rate, pause rate, and short-time energy) and the evaluated emotions is determined, which enhances previous similar studies that have no...
EURASIP Journal on Image and Video Processing, 2018
We propose the first non-invasive three-layer architecture in literature based on neural networks... more We propose the first non-invasive three-layer architecture in literature based on neural networks that aims to determine the Big Five personality traits of an individual by analyzing offline handwriting. We also present the first database in literature that links the Big Five personality type with the handwriting features collected from 128 subjects containing both predefined and random texts. Testing our novel architecture on this database, we show that the predefined texts add more value if enforced on writers in the training stage, offering accuracies of 84.4% in intra-subject tests and 80.5% in inter-subject tests when the random dataset is used for testing purposes, up to 7% higher than when random datasets are used in the training phase. We obtain the highest prediction accuracy for Openness to Experience, Extraversion, and Neuroticism (over 84%), while for Conscientiousness and Agreeableness, the prediction accuracy is around 77%. Overall, our approach offers the highest accuracy compared with other state-of-the-art methods and results are computed in maximum 90 s, making the approach faster than the questionnaire or psychological interviews currently used for determining the Big Five personality traits. Our research also shows there are relationships between specific handwriting features and prediction with high accuracy of specific personality traits and this can be further exploited for improving, even more, the prediction accuracy of the proposed architecture.
EURASIP Journal on Image and Video Processing, 2017
We propose a novel three-layered neural network-based architecture for predicting the Sixteen Per... more We propose a novel three-layered neural network-based architecture for predicting the Sixteen Personality Factors from facial features analyzed using Facial Action Coding System. The proposed architecture is built on three layers: a base layer where the facial features are extracted from each video frame using a multi-state face model and the intensity levels of 27 Action Units (AUs) are computed, an intermediary level where an AU activity map is built containing all AUs' intensity levels fetched from the base layer in a frame-by-frame manner, and a top layer consisting of 16 feed-forward neural networks trained via backpropagation which analyze the patterns in the AU activity map and compute scores from 1 to 10, predicting each of the 16 personality traits. We show that the proposed architecture predicts with an accuracy of over 80%: warmth, emotional stability, liveliness, social boldness, sensitivity, vigilance, and tension. We also show there is a significant relationship between the emotions elicited to the analyzed subjects and high prediction accuracy obtained for each of the 16 personality traits as well as notable correlations between distinct sets of AUs present at high-intensity levels and increased personality trait prediction accuracy. The system converges to a stable result in no more than 1 min, making it faster and more practical than the Sixteen Personality Factors Questionnaire and suitable for real-time monitoring of people's personality traits.
Study on determining the Big-Five personality traits of an individual based on facial expressions
2015 E-Health and Bioengineering Conference (EHB), 2015
Previous studies revealed an increasing interest in determining the personality and behavior of i... more Previous studies revealed an increasing interest in determining the personality and behavior of individuals in areas such as career development and counseling, personalized health assistance, mental disorder diagnosis as well as detection of physical diseases with personality shift symptoms. Current ways of determining the Big-Five personality types involve completing a questionnaire, that takes an impractical amount of time and it cannot be used often. Our research aims building a novel non-invasive system to determine Big-Five personality traits based on facial features acquired using Facial Action Coding System. Results show links between the FACS action units present at maximum intensities in facial features and the personality traits of the individual. Moreover, the system built offers over 75% accuracy in predicting openness to experience, as well as neuroticism and extraversion and proves practical, offering results in no more than 3 minutes compared to the amount of time taken to complete a questionnaire.
Study on determining the Myers-Briggs personality type based on individual's handwriting
2015 E-Health and Bioengineering Conference (EHB), 2015
Studies in psychology showed a close link between handwriting and personality, but this was never... more Studies in psychology showed a close link between handwriting and personality, but this was never formally analyzed. In the context of career development there is a need to determine the personality type in a more efficient manner than the classic questionnaire. Moreover, in the fields of psychology and medicine, constant monitoring the patient's personality can provide information regarding his mental health status, if he suffers from mental disorders or show psychological symptoms for common physical diseases. We analyze the link between personality types and handwriting, by correlating the handwriting features with the personality primitives in a neural-network 3-level architecture. Results show an accuracy of 86.7% in determining the personality type, with highest accuracies for Extravert vs. Introvert and Thinking vs. Feeling personality primitives. The system computes the personality type in less than 1 minute, proving to be more efficient than a questionnaire and suitable for real-life use.
Recognizing human gestures in videos by modeling the mutual context of body position and hands movement
Multimedia Systems, 2016
In our days, due the evolution of high-speed computers, the old Human–Computer Interface (HCI) le... more In our days, due the evolution of high-speed computers, the old Human–Computer Interface (HCI) legacies based on mouse and keyboard are slowly becoming obsolete and cannot be accurate enough and respond in a timely manner to the flow of information today. This is why new ways of communicating with the computer have to be researched, the most natural one being the use of gestures. In this paper, a two-level architecture for recognizing human gestures from video frames is proposed. The architecture makes use of several feed-forward neural networks to compute the gestures based on the Haar-like features of body, hand and finger as well as a stochastic-free context grammar that is employed to comprise the mutual context between body pose and hand movement. Trained and tested on 10 gestures (Swipe Right, Swipe Left, Swipe Up, Swipe Down, Horizontal Wave, Vertical Wave, Circle, Point, Palm Up and Fist) the over 94 % accuracy of the system surpasses the current state of the art and compared with a system with no mutual context between body position and hand movement our proposed architecture shows an increase in accuracy with up to 7 %.
IET Biometrics, 2016
Although current face recognition systems in biometrics field are accurate enough to be used as s... more Although current face recognition systems in biometrics field are accurate enough to be used as substitutes for passwords or keys, most of them are prone to face spoofing attacks. Different techniques for face spoofing identification have been researched but most of them introduce additional sensors and are not cost or computationally efficient. In this paper we study the possibility of using individual differences in facial expressions for improving a face recognition system and make it immune to spoofing attacks. We develop a soft biometric neural network based system for video-based face recognition by analysing patterns in individual facial expressions on multiple frames. Results show that such a system is possible and has accuracies higher than 85%. Used alongside with a standard PCA-based face recognition system, the combined method achieved 94.5% accuracy on Honda/UCSD Video Database and 92.9% on Youtube Faces DB, comparable with state-of-the-art. When tested against photo spoofing attacks on three public anti-spoofing databases the proposed method was immune. In terms of video spoofing, the error rate for our proposed method was 1% surpassing state-of-the-art methods. IET Review Copy Only IET Biometrics This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.
Recognizing emotions from videos by studying facial expressions, body postures and hand gestures
2015 23rd Telecommunications Forum Telfor (TELFOR), 2015
A system for recognizing emotions from videos by studying facial expressions, hand gestures and b... more A system for recognizing emotions from videos by studying facial expressions, hand gestures and body postures is presented. A stochastic context-free grammar (SCFG) containing 8 combinations of hand gestures and body postures for each emotion is used and we show that increasing the number of combinations in SCFG improves the system's generalization for new hand gesture and body posture combinations. We show that hand gestures and body postures contribute to improving the emotion recognition rate with up to 5% for Anger, Sadness and Fear compared to the standard facial emotion recognition system, while for Happiness, Surprise and Disgust no significant improvement was noticed.
Noise robust Automatic Speech Recognition system by integrating Robust Principal Component Analysis (RPCA) and Exemplar-based Sparse Representation
2015 7th International Conference on Electronics, Computers and Artificial Intelligence (ECAI), 2015
An enhanced Automatic Speech Recognition (ASR) system based on Hidden Markov Models (HMM) is pres... more An enhanced Automatic Speech Recognition (ASR) system based on Hidden Markov Models (HMM) is presented. The system makes use of two sparse algorithms in order to remove the noise from the speech signal and improve the overall ASR recognition rate: Robust Principal Component Analysis (RPCA) and Exemplar-based Sparse Representation. We start with the premise that RPCA offers better results at lower Signal-to-noise ratios (SNRs) while Exemplar-based Sparse Representation offers good results for SNRs lower than 15 dB, and therefore we envisage architecture able to select between the two algorithms depending on the SNR detected in the speech signal. We present the architecture of our proposed model, as well as the experimental results in different scenarios and the improvements that can be brought in future researches.
Improved automatic speech recognition system using sparse decomposition by basis pursuit with deep rectifier neural networks and compressed sensing recomposition of speech signals
2014 10th International Conference on Communications (COMM), 2014
Research on the common limitations of Automatic Speech Recognition (ASR) systems state problems r... more Research on the common limitations of Automatic Speech Recognition (ASR) systems state problems ranging from environmental noise, and channel or speaker variability to the limitations imposed by the measurement device. In mobile applications for automatic speech recognition, the Nyquist criteria imposes more limitations on the sampling rate at which a device can acquire signal, often the lack of fidelity of the acquired signal causing bad speech recognition. This is a specific problem for mobile devices (which are also, nowadays, the prime beneficiaries of speech recognition applications) as in this case the sampling rate is limited. We envisage a way to get the best out of any acquired signal, by use of sparsity decomposition algorithms and compressed sensing recomposition. We focus on the fact that complex sounds can be viewed as an overlapping of a number of sounds coming from simple sparse sources. Therefore, we decompose the measured signal in a linear combination of simple sparse signals and we reconstruct each sparse signal by means of compressed sensing recomposition in order to gain a better signal fidelity. We make use of deep rectifier neural network designed to decompose a training set of signals and compute a specific dictionary with simple sparse signals. The resulted sparse signals are used for decomposing the acquired signal by means of sparse algorithms, and, consequently, the resulted combination of sparse signals will be used for signal reconstruction in a compressed sensing algorithm. We test the framework for different simulated speech signals, as well as its usability in automatic speech recognition, discussing the improvements this upgrade brings to an ASR. In this paper we will describe the framework and the algorithms used and present the experimental results.
Proposed architecture of a fully integrated modular neural network-based automatic facial emotion recognition system based on Facial Action Coding System
2014 10th International Conference on Communications (COMM), 2014
In this paper we describe the architecture of a fully integrated modular neural network-based aut... more In this paper we describe the architecture of a fully integrated modular neural network-based automatic facial emotion recognition (FER) system able to recognize emotions based on the Facial Action Code System (FACS). The proposed framework makes use of a neural network to combine the recognition results from different sources, improving the integration of different types of classifiers, in order to provide better facial emotion recognition results. We present the architecture and the implementation details, as well as results and possible improvements that can be brought to the current framework.
Study on determining the Myers-Briggs personality type based on individual's handwriting
Print
Request Permissions
Studies in psychology showed a close link between handwriting and pers... more Print
Request Permissions
Studies in psychology showed a close link between handwriting and personality, but this was never formally analyzed. In the context of career development there is a need to determine the personality type in a more efficient manner than the classic questionnaire. Moreover, in the fields of psychology and medicine, constant monitoring the patient's personality can provide information regarding his mental health status, if he suffers from mental disorders or show psychological symptoms for common physical diseases. We analyze the link between personality types and handwriting, by correlating the handwriting features with the personality primitives in a neural-network 3-level architecture. Results show an accuracy of 86.7% in determining the personality type, with highest accuracies for Extravert vs. Introvert and Thinking vs. Feeling personality primitives. The system computes the personality type in less than 1 minute, proving to be more efficient than a questionnaire and suitable for real-life use.
Study on determining the Big-Five personality traits of an individual based on facial expressions
Previous studies revealed an increasing interest in determining the personality and behavior of i... more Previous studies revealed an increasing interest in determining the personality and behavior of individuals in areas such as career development and counseling, personalized health assistance, mental disorder diagnosis as well as detection of physical diseases with personality shift symptoms. Current ways of determining the Big-Five personality types involve completing a questionnaire, that takes an impractical amount of time and it cannot be used often. Our research aims building a novel non-invasive system to determine Big-Five personality traits based on facial features acquired using Facial Action Coding System. Results show links between the FACS action units present at maximum intensities in facial features and the personality traits of the individual. Moreover, the system built offers over 75% accuracy in predicting openness to experience, as well as neuroticism and extraversion and proves practical, offering results in no more than 3 minutes compared to the amount of time taken to complete a questionnaire.
Recognizing emotions from videos by studying facial expressions, body postures and hand gestures
A system for recognizing emotions from videos by studying facial expressions, hand gestures and b... more A system for recognizing emotions from videos by studying facial expressions, hand gestures and body postures is presented. A stochastic context-free grammar (SCFG) containing 8 combinations of hand gestures and body postures for each emotion is used and we show that increasing the number of combinations in SCFG improves the system's generalization for new hand gesture and body posture combinations. We show that hand gestures and body postures contribute to improving the emotion recognition rate with up to 5% for Anger, Sadness and Fear compared to the standard facial emotion recognition system, while for Happiness, Surprise and Disgust no significant improvement was noticed.
Noise robust Automatic Speech Recognition system by integrating Robust Principal Component Analysis (RPCA) and Exemplar-based Sparse Representation
An enhanced Automatic Speech Recognition (ASR) system based on Hidden Markov Models (HMM) is pres... more An enhanced Automatic Speech Recognition (ASR) system based on Hidden Markov Models (HMM) is presented. The system makes use of two sparse algorithms in order to remove the noise from the speech signal and improve the overall ASR recognition rate: Robust Principal Component Analysis (RPCA) and Exemplar-based Sparse Representation. We start with the premise that RPCA offers better results at lower Signal-to-noise ratios (SNRs) while Exemplar-based Sparse Representation offers good results for SNRs lower than 15 dB, and therefore we envisage architecture able to select between the two algorithms depending on the SNR detected in the speech signal. We present the architecture of our proposed model, as well as the experimental results in different scenarios and the improvements that can be brought in future researches.
Improved Automatic Speech Recognition system by using compressed sensing signal reconstruction based on L0 and L1 estimation algorithms
This paper presents a way of improving the recognition rate of a typical Hidden Markov Model (HMM... more This paper presents a way of improving the recognition rate of a typical Hidden Markov Model (HMM) -based Automatic Speech Recognition (ASR) system by integrating the l1 - least absolute deviation (LAD) algorithm and the l0 - least square (LS) algorithm in a framework designed to selectively use them based on the level of impulse noise present in speech signal. We present the overall architecture of the model, as well as experimental results and compare our enhanced noise-robust HMM-based ASR system with state-of-the-art proving the improvements brought by this approach as well as future directions of research.
Uploads
Papers by Mihai Gavrilescu
Request Permissions
Studies in psychology showed a close link between handwriting and personality, but this was never formally analyzed. In the context of career development there is a need to determine the personality type in a more efficient manner than the classic questionnaire. Moreover, in the fields of psychology and medicine, constant monitoring the patient's personality can provide information regarding his mental health status, if he suffers from mental disorders or show psychological symptoms for common physical diseases. We analyze the link between personality types and handwriting, by correlating the handwriting features with the personality primitives in a neural-network 3-level architecture. Results show an accuracy of 86.7% in determining the personality type, with highest accuracies for Extravert vs. Introvert and Thinking vs. Feeling personality primitives. The system computes the personality type in less than 1 minute, proving to be more efficient than a questionnaire and suitable for real-life use.