2019/05/01 by Jennifer Sorinasa, Juan C. Fernandez-Troyano, Sorinasa, Jennifer +7
Neuroscience · Medicine · Psychology · #EEG and Brain-Computer Interfaces #ECG Monitoring and Analysis #Emotion and Mood Recognition
paper · pdf · doi:10.48550/arxiv.1905.00230
The large range of potential applications, not only for patients but also for\nhealthy people, that could be achieved by affective BCI (aBCI) makes more\nlatent the necessity of finding a commonly accepted protocol for real-time\nEEG-based emotion recognition. Based on wavelet package for spectral feature\nextraction, attending to the nature of the EEG signal, we have specified some\nof the main parameters needed for the implementation of robust positive and\nnegative emotion classification. 12 seconds has resulted as the most\nappropriate sliding window size; from that, a set of 20 target\nfrequency-location variables have been proposed as the most relevant features\nthat carry the emotional information. Lastly, QDA and KNN classifiers and\npopulation rating criterion for stimuli labeling have been suggested as the\nmost suitable approaches for EEG-base emotion recognition. The proposed model\nreached a mean accuracy of 98% (s.d. 1.4) and 98.96% (s.d. 1.28) in a\nsubject-dependent approach for QDA and KNN classifier, respectively. This new\nmodel represents a step forward towards real-time classification. Moreover,\nalthough results were not conclusive, new insights regarding\nsubject-independent approximation have been discussed.\n