Reshma Kar

Detection of signaling pathways in human brain during arousal of specific emotion

Neuroscientists usually determine similarity between EEG electrode signals, by a measure of pairwise linear dependence among them. However, recent research indicates the drawbacks of analyzing the pairwise dependence of signals instead of analyzing the simultaneous joint interdependence among them. To overcome this problem we propose a novel similarity measure known as probabilistic relative correlation. Our approach is unique because our similarity measure allows the electrodes to have probabilistic similarity measures and recognizes emotion dependent structures even from mismatched sequences of correlation. We further validate our proposed similarity measure by testing it on the well-known emotion recognition problem. Our experiments have noteworthy implications towards realizing the neural signatures of discrete emotions and will allow for the better understanding of neurological pathways associated with different emotional states. To identify the most active neurological pathways in brain during an emotion, we adapt the minimal spanning tree algorithm.

Emotion Recognition System by Gesture Analysis Using Fuzzy Sets

Gestures have been called the leaky source of emotional information. Also gestures are easy to retrieve from a distance by ordinary cameras. Thus as many would agree gestures become an important clue to the emotional state of a person. In this paper we have worked on recognizing emotions of a person by analyzing only gestural information. Subjects are initially trained to perform emotionally expressive gestures by a professional actor. The same actor trained the system to recognize the emotional context of gestures. Finally the gestural performances of the subjects are evaluated by the system to identify the class of emotion indicated. Our system yields an accuracy of 94.4% with a training set of only one gesture per emotion. Apart from this our system is also computationally efficient. Our work analyses emotions from only gestures, which is a significant step towards reducing the cost efficiency of emotion recognition. It may be noted here that this system may also be used for the purpose of general gesture recognition. We have proposed new features and a new classifying approach using fuzzy sets. We have achieved state of art accuracy with minimal complexity as each motion trajectory along each axis generates only 4 displacement features. Each axis generates a trajectory and only 6 joint trajectories among all joint trajectories are compared. The 6 motion trajectories are selected based on maximum motion, as maximum moving regions give more information on gestures. The experiments have been performed on data obtained from Microsoft Kinect sensors. Training and Testing were subject gender independent.

Dance Composition Using Microsoft Kinect

In this work, we propose a novel approach in which a system autonomously composes dance sequences from previously taught dance moves with the help of the well-known differential evolution algorithm. Initially, we generated a large population of dance sequences. The fitness of each of these sequences was determined by calculating the total inter-move transition abruptness of the adjacent dance moves. The transition abruptness was calculated as the difference of corresponding slopes formed by connected body joint co-ordinates. By visually evaluating the dance sequences created, it was observed that the fittest dance sequence had the least abrupt inter-move transitions. Computer simulation undertaken revealed that the developed dance video frames do not have significant inter-move transition abruptness between two successive frames, indicating the efficacy of the proposed approach. Gestural data specific of dance moves is captured using a Microsoft Kinect sensor. The algorithm developed by us was used to fuse the dancing styles of various ‘Odissi’ dancers dancing to the same rasa (theme) and tala (beats) and loy (rhythm). In future, it may be used to fuse different forms of dance.

EEG source localization by memory network analysis of subjects engaged in perceiving emotions from facial expressions

The memory network is a result of current dipoles created in the brain. Localizing the source of these current flows is known as source localization, and it could potentially reveal which parts of the brain are actually responsible for a particular brain activity. It would also increase the spatial resolution of an EEG recording by identifying the true source of multiple correlated readings. In our experiments, we employed memory networks to classify perception of emotional instances conveyed in facial expressions as well as to localize sources. These networks were created by selectively evaluating EEG channel signals pairwise for Granger causality. Channel selection was based on clustering of EEG features by Self Organizing Feature Map (SOFM). Principal Component Analysis (PCA) was employed for dimension reduction and noise elimination of EEG features. Finally a new metric based on Fischers discriminant was used to compare different source localization techniques, where real source locations are unknown. The perception of the stimuli was classified as belonging to one the following classes i) Happy ii) Sad iii) Fear iv) Relaxed. The created memory networks could classify perception of emotional content in 90.64% of cases. Comparison by the proposed Fischer Discriminant based metric revealed that the proposed network identification technique performs better at source localization as compared to independent component based source localization.

Extending the Nelder-Mead algorithm for feature selection from brain networks

Centrifugation is often applied in laboratories and industries to increase the effective gravity on a particle and hence make it sediment faster. Based on this principle, one may extend the existing optimization techniques, which are driven by only gravitational force (objective values of discovered best solutions), and do not consider application of centrifugal force for faster convergence. We extended the Nelder-Meads simplex algorithm, by applying an exponentially decaying centrifugal force on each of the computed vertices of the simplex. The proposed centrifugation technique was also applied on other optimization algorithms including differential evolution and gravitational search algorithms. It was seen that application of centrifugal force indeed enhanced the objective values obtained by of all the tested evolutionary algorithms. The comparative performance of the extended Nelder-Mead Algorithm was found to be better among all the tested algorithms. The algorithms were compared on the basis of the best obtained objective value after a fixed number of objective function evaluations (here 20 times the problem dimension). Testing was performed in the real world problem of EEG feature selection (from brain networks), for the classification of memory encoding versus recall using SVM. The average classification accuracy was found to be high (89.97%).

Joint pain detection by gait analysis for elderly healthcare

This Lower part of skeletal structure of humans (hereafter, lower body) plays a fundamental role in maintaining balance and gait of a person. Naturally, when the lower body joints are affected by pain, the gait of a person is changed. In this paper we inspected the correspondence of gait patterns to different lower body joint pains. It was found that gait patterns can indeed be used as non-intrusive bio-markers for detecting selective body joint pains. From the computational point of view, our work has three major standpoints. First, we used a well-established statistical technique known as differencing to enable body structure and location independent recognition of gait. Second we proposed a scheme of selecting body joints which are most informative of body pain by assigning maximum score to body joint trajectories which are unique to different classes of joint pain but similar for the same class of joint pain. Third, we employ differential evolution algorithm for compression of motion trajectories by a ratio of 1:5 by selecting only important samples which maximize the entropy of the motion trajectory. We apply the above three techniques for data enhancement and use it as a feature extraction technique. It is a well-established fact that efficient classification is dependent on judicious selection of feature-space data which captures essential information hidden in the data while minimizing its dimensionality. Thus our approach of data enhancement by the above steps outperforms an established traditional feature extraction scheme by a wide margin. The average classification accuracy obtained by us is 87.91%.

A general type-2 fuzzy set induced single trial P300 detection

P300 is one of the most widely studied event-related potentials. Unfortunately, most of the existing automatic P300 detection schemes require computations over repetitive trials in both training and recognition phases. Several attempts have recently been endeavored towards the single trial detection of the P300 signals. However, no acceptable solution to the problem is found till date. In the present work, we have attempted to address this problem in the light of latency and (amplitude) deflection of the signal. The intra- and inter-personal variations inherent in these features are managed by the uncertainty management characteristics of General Type-2 Fuzzy Sets. First, these sets are constructed by exploiting the knowledge obtained from different trials of a large number of subjects. The secondary membership functions of the Type-2 Fuzzy Sets are computed based on a novel density dependent measure of the primary membership functions in the footprint of uncertainty. Second, recognition of P300 in an unknown EEG trial is performed based on the agreement of measured feature values with the General Type-2 Fuzzy knowledge-base. Majority voting of the concerned electrodes makes the scheme more robust. The experimental results show that the proposed algorithm is capable of achieving 88.60% accuracy in single trial detection of P300 instances, which is significantly higher than those obtained in state of the art algorithms.

Fuzzy rule enhanced support vector machines for classification of emotions from brain networks

Support Vector Machines are widely accepted in the field of pattern recognition because of their superiority in performing supervised classification. It is known that all kernel parameters may be used for classification more-or-less precisely (giving rise to vagueness) and also for the same classification problem, there are a number of kernel parameters which give the best accuracy (giving rise to uncertainty). Hence, an appropriate scheme of representing best suited kernel parameters for a given classification problem requires an Interval-type 2 approach. In this work the authors introduce a fuzzy rule-based kernel parameter selection technique which is based on the variability (inter-class and intra-class scatter) of the dataset to be classified. A significant advantage of using the proposed fuzzy kernel parameter selection technique is that one can identify the kernel parameter which has least curvature and hence avoid over fitting. The introduced method of kernel parameter selection is tested in an emotion recognition problem by brain network analysis. Experiments undertaken indicate that selection of appropriate kernel parameters can increase accuracy up to 30%.

A novel feature extraction scheme for N400 detection

The presented work proposes a simple feature extraction technique which is designed for robust detection of event related potentials (ERP). This technique was tested to detect the N400 which is an ERP generally associated with recall. The chief advantages of the proposed technique are that it is robust to different ocular artifacts and yet sensitive to event related potentials. Further each signal will correspond to only a few features as opposed to 100s and 1000s of features obtained by traditional feature extraction techniques. The proposed steps involve a) Computing the first and second order difference of the data b) measuring mean and variance respectively for first and second order differencing over 1 second windows c) repeating the steps a and b after lagging the signal by 0.5 seconds. Differencing computes the change in amplitude of EEG signals, which is considered important in ERP analysis. Step (b) is a unique way of getting rid of abrupt signal changes which are artifacts, as for abrupt changes in the signal; the computed variance of the second difference is high. Also, computing windowed average of the first difference reveals the (increasing/ decreasing) trend of the data. Step (c) ensures that potential changes are not missed if they lie across two windows during the first phase of windowing. The proposed approach of feature extraction by the above steps outperforms three established traditional feature extraction schemes in identifying N400 waveforms using support vector machines. The average classification accuracy obtained by the proposed feature set is 96.91%.

An Evolutionary Approach to Secondary Membership Function Selection in Generalized Type-2 Fuzzy Sets

Lack of knowledge about secondary membership function acts as an impediment to using generalized type-2 fuzzy sets in real-world problems. This chapter shows a new direction to compute secondary memberships in the settings of a strategic optimization problem. It employs three strategies to design an optimization objective as a function of secondary memberships and employs differential evolution algorithm to determine secondary memberships as the optimal solution to the optimization problem. The proposed method of secondary membership function evaluation has successfully been applied to an emotion recognition problem.