An Ensemble Learning Method for Emotion Charting Using Multimodal Physiological Signals.

Area of Science:

• Physiological signal processing
• Machine learning for emotion recognition
• Neuroscience and affective computing

Background:

• Emotion charting relies on multimodal physiological signals (EEG, ECG, GSR) for applications in healthcare and neuromarketing.
• Accurate classification of these autonomous signals is challenging due to their non-linear nature and recording noise.
• Existing methods struggle with precise emotion detection, necessitating advanced classification techniques.

Purpose of the Study:

• To present a novel ensemble learning method using deep learning for classifying four distinct emotional states (HVHA, LVLA, HVLA, LVHA).
• To enhance the accuracy of emotion charting by effectively processing multimodal physiological signals.
• To address the challenges posed by signal noise and non-linearity in emotion detection.

Main Methods:

• Multimodal signals (EEG, ECG, GSR) were preprocessed using bandpass filtering and Independent Component Analysis (ICA).
• Discrete Wavelet Transform (DWT) was applied for time-frequency conversion, generating signal spectrograms.
• Features were extracted using Stacked Autoencoders, followed by classification with SVM, RF, and LSTM, combined via majority voting for ensemble classification.

Main Results:

• The proposed deep learning ensemble method achieved a highest accuracy of 94.5% on the AMIGOS dataset.
• The system demonstrated improved performance compared to other state-of-the-art methods for emotion classification.
• Feature extraction from spectrograms using autoencoders proved effective for emotion identification.

Conclusions:

• The novel ensemble learning approach offers a robust and accurate solution for emotion charting using physiological signals.
• This method effectively mitigates noise and non-linearity issues inherent in physiological data.
• The findings highlight the potential of deep learning and ensemble techniques in advancing affective computing and emotion recognition.