Performance evaluation of multi-channel electroencephalogram signal (EEG) based time frequency analysis for human emotion recognition. (September 2022)

• Record Type: Journal Article
• Authors: Wagh, Kalyani P.; Vasanth, K.
• Is Part Of: Biomedical signal processing and control. Volume 78(2022)
• Publication Date: 2022-09
• Subjects: EEG -- Emotion -- Discrete Wavelet Transform; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.103966
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: The EEG-based approach is an effective mechanism that is extensively utilized for emotion identification in real-world settings. DWT is used to decompose signal in various frequency bands with "db6" to get features like PSD, Energy, Std. Dev., Variance. Simultaneously we derived time domain features Hjorth parameters, Max. and Min. value, Kurtosis, Skewness from EEG signal. The work is carried out with five electrode pairs FP1-FP2, F3-F4, T3-T4, P7-P8 and O1-O2 based on various brain regions. Abstract: The automated detection of a human's emotional state by acquiring physiological or non-physiological cues is referred to as Emotion Recognition. The EEG-based approach is an effective mechanism that is extensively utilized for emotion identification in real-world settings. In this paper various classifiers are used to classify the EEG signal into three emotional states using SEED database, prepare for emotion study using physiological signals. DWT is used to decompose EEG signal in various frequency bands with "db6" as wavelet function for deriving various features like PSD, Energy, Standard Deviation, and Variance. Simultaneously we have derived various time domain features like Hjorth parameters, Maximum and Minimum value, Kurtosis, Skewness from EEG signal for recognition of emotion. The work is carried out with five electrode pairs Prefrontal (FP1-FP2), Frontal (F3-F4), Temporal (T3-T4), Parietal (P7-P8) and Occipital (O1-O2) out of 62 electrodes. Three classification methods, Support Vector Machine, K Nearest Neighbor and Decision Tree are used and their performances are compared for categorizing emotional state. The trial results show that maximum classification rate is 71.52% using decision tree and 60.19% using KNN. Electrodes FP1 and FP2 performs well in classification. Higher frequency spectrum like gamma and beta performs well in emotion recognition.