Innovative Poincare's plot asymmetry descriptors for EEG emotion recognition

Given the importance of emotion recognition in both medical and non-medical applications, designing an automatic system has captured the attention of several scholars. Currently, EEG-based emotion recognition has a special position, which has not fulfilled the desired accuracy rates yet. This experiment intended to provide novel EEG asymmetry measures to improve emotion recognition rates. Four emotional states have been classified using the k-nearest neighbor (kNN), support vector machine, and Naïve Bayes. Feature selection has been performed, and the role of employing a different number of top-ranked features on emotion recognition rates has been assessed. To validate the efficiency of the proposed scheme, two public databases, including the SJTU Emotion EEG Dataset-IV (SEED-IV) and a Database for Emotion Analysis using Physiological signals (DEAP) were evaluated. The experimental results indicated that kNN outperformed the other classifiers with a maximum accuracy of 95.49 and 98.63% using SEED-IV and DEAP datasets, respectively. In conclusion, the results of the proposed novel EEG-asymmetry measures make the framework a superior one compared to the state-of-art EEG emotion recognition approaches.

A Multiclass Arousal Recognition using HRV Nonlinear Analysis and Affective Images

This paper reports on a multiclass arousal recognition system based on autonomic nervous system linear and nonlinear dynamics during affective visual elicitation. We propose a new hybrid method based on Lagged Poincaré Plot (LPP) and symbolic analysis, hereinafter called LPP_symb. This tool uses symbolic analysis to evaluate the irregularity of the trends of Lagged Poincaré Plot (LPP) quantifiers over the lags, and is here applied to investigate complex Heart Rate Variability (HRV) changes during emotion stimuli. In the experimental protocol 22 healthy subjects were elicited through a passive visualization of affective images gathered from the international affective picture system. LPP_symb and standard HRV analysis (defined in time and frequency domains) were applied to HRV series of one minute length. Then, an ad-hoc pattern recognition algorithm based on quadratic discriminant classifier was implemented and validated through a leave-onesubject-out procedure. The best performance of the proposed classification algorithm for recognizing the four classes of arousal was obtained using nine features comprising heartbeat complex dynamics, achieving an accuracy of 71.59%.

A NOVEL SIGNAL-BASED FUSION APPROACH FOR ACCURATE MUSIC EMOTION RECOGNITION

The objective of this study is to propose an accurate emotion recognition methodology. To this end, a novel fusion framework based on wavelet transform (WT), and matching pursuit (MP) algorithm was offered. Electrocardiogram (ECG) and galvanic skin response (GSR) of 11 healthy students were collected while subjects listened to emotional music clips. In both fusion techniques, Coiflet wavelet (Coif5 at level 14) was chosen as a wavelet family and MP dictionary, respectively. After employing the proposed fusion framework, some statistical measures were extracted. To describe emotions, three schemes were adopted: two-dimensional model (five classes), valence-(three classes), and arousal-(three classes) based emotion categories. Subsequently, the probabilistic neural network (PNN) was applied to classify affective states. The experiments indicate that the MP-based fusion approach outperform the wavelet-based fusion technique or methods using only ECG or GSR indices. Considering the proposed fusion techniques, the maximum classification rate of 99.64% and 92.31% was reached for the fusion methodology based on the MP algorithm (five classes of emotion) and wavelet-based fusion technique (three classes of valence), respectively.