Emotion Recognition from ECG Signals Contaminated by Motion Artifacts

Emotion recognition (ER) using Electrocardiogram (ECG) has drawn increasing attention with the rapid development of inexpensive and wearable ECG devices. ER using ECG signals contaminated by motion artifacts (MA) is a difficult problem, since MA may lead to the decline of the distinguish ability of ECG features and then make the performance of an emotion recognition model degrade. So far few work has studied this problem. In this study, a method is proposed for ER using ECG signals contaminated by MA, which consists of singular spectrum analysis for removing MA in ECG signals, feature extraction and fusion for extracting features from denoised ECG signals, and emotion classification used to recognize emotions. Experimental results on ECG signals containing MA showed that the performance of ER may degrade due to MA, and that the proposed method has a high classification accuracy even when ECG signals contain MA.