Evaluation of feature extraction techniques in emotional state recognition

Author

Bastos-Filho, T.F. ; Ferreira, Andre ; Atencio, A.C. ; Arjunan, S. ; Kumar, Dinesh

Author_Institution

Fed. Univ. of Espirito Santo, Vitoria, Brazil

fYear

2012

Firstpage

1

Lastpage

6

Abstract

We present in this paper a study of three EEG signals feature extraction techniques. These techniques have been widely employed in researches of emotional states recognition: statistical characteristics, features based on PSD (Power Spectral Density) and features based on HOC (High Order Crossings). The validation was performed via classification of emotional states of calm and stress using the K-NN based classifier in off-line mode using EEG signals from available DEAP database. The best results achieved were 70.1%, using the PSD based technique, and 69.59% using the HOC based technique.

Keywords

electroencephalography; emotion recognition; feature extraction; neurophysiology; pattern classification; state estimation; statistical analysis; DEAP database; EEG signal feature extraction techniques; HOC based features; HOC based technique; K-NN based classifier; PSD based features; emotional state classification; emotional state recognition; feature extraction technique evaluation; high order crossings; off-line mode; power spectral density; statistical characteristics; Brain modeling; Databases; Electroencephalography; Emotion recognition; Feature extraction; Stress; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Intelligent Human Computer Interaction (IHCI), 2012 4th International Conference on

Conference_Location

Kharagpur

Print_ISBN

978-1-4673-4367-1

Type

conf

DOI

10.1109/IHCI.2012.6481860

Filename

6481860