A comparison of emotion recognition system using electrocardiogram (ECG) and photoplethysmogram (PPG)

Electrocardiogram (ECG) and Photoplethysmogram (PPG) are derived from electrical signals of the heart activities and frequently used to diagnose and monitor cardiovascular disease. In the field of affective computing, these two signals can be used to recognize human emotions, this is supported by the wide availability of wearable devices that able to collect ECG or PPG in the market. ECG is frequently used as a unimodal signal for ERS, but the usage of PPG signals as unimodal ERS is still limited. There is no consensus about whether ECG is more suitable than PPG in ERS or vice versa. Only a few research have compared ECG and PPG. Therefore, this work intends to close this gap by developing an ERS employing ECG and PPG and evaluating the efficacy of both signals in ERS. This is done through data collected from 47 participants and two public datasets. The result from the data collected indicates that ECG is superior at recognizing arousal emotion with accuracy up to 68.75%, whereas PPG superior at recognizing valence up to 64.94% and dimension classes with 37.01% accuracy. The findings suggest that despite the current trend where researchers favour ECG over PPG, the PPG signals can be used as the only modality in developing ERS with results comparable to those obtained using ECG signals.