Physiological Signal Processing for Emotional Feature Extraction

This paper introduces new approaches of physiological signal processing prior to feature extraction from electrocardiogram (ECG) and electromyography (EMG). Firstly a new signal denoising approach based on the Empirical mode decomposition (EMD) is presented. The EMD can decompose the noisy signal into a number of Intrinsic Mode Functions (IMFs). The proposed algorithm estimates the noise level of each IMF. Experiments show that the proposed EMD-based method provides better denoising results compared to state-of-art. In addition, a real-time QRS detection approach is proposed to be directly applied on the noisy ECG signals. Moreover, an adaptive thresholding approach is employed for the EMG segmentation. Both approaches are validated using synthetic and real physiological data resulting in good performances.