Emotion Recognition Using Spectral Feature from Facial Electromygraphy Signals for Human-Machine Interface.

Area of Science:

• Biomedical Engineering
• Affective Computing
• Signal Processing

Background:

• Emotion recognition is vital for healthcare and human-machine interaction.
• Facial electromyography (EMG) signals offer insights into emotional states.
• Current methods require further enhancement for robust emotion classification.

Purpose of the Study:

• To classify emotions using a spectral feature from facial EMG signals in the valence dimension.
• To evaluate the efficacy of peak frequency values as a classification feature.
• To explore the utility of this method in myoelectric control.

Main Methods:

• Facial EMG signals were acquired from the DEAP dataset.
• Short-Time Fourier Transform was applied to analyze the signals.
• Peak frequency values were extracted per second and classified using a Support Vector Machine (SVM).

Main Results:

• The extracted spectral feature achieved 61.37% accuracy in classifying valence emotions.
• SVM classification demonstrated the potential of the proposed feature.
• The method shows promise for enhancing emotion recognition systems.

Conclusions:

• A novel spectral feature from facial EMG can classify valence emotions with moderate accuracy.
• This feature can serve as a valuable addition to existing emotion recognition techniques.
• The analysis method holds potential for applications in myoelectric control.