Search

It has been well established that change in emotion state is associated with the change in physiological signals. This paper aimed to investigate the differences of finger photoplethysmography (PPG) morphological features and feature time series between happiness and sadness emotion states. Fifty-three volunteers were enrolled. Finger PPG signals were recorded under two emotion states with a random measurement order (first happiness emotion measurement then sadness or reverse). Seven morphological features were extracted, including three temporal features (T, T1 and T2), three area features (A, A1 and A2) and one amplitude feature (Amp). Five variability indices from the 5-min feature time series were calculated, including two time-domain indices (SDNN and RMSSD) and three frequency-domain indices (LFn, HFn and LF/HF). Results showed that temporal features T2 and T were critical features for identifying the two emotion states since not only they themselves but also their three frequency-domain variability indices had significant differences between the two emotion states. For area features, only two frequency-domain variability indices of LFn and HFn for A1 feature time series reported significant differences. Amplitude feature Amp itself, as well as its variability indices, did not had significant differences between the two emotion states. These results indicated that temporal features were more sensitive to response to emotion change than area and amplitude features. Moreover, compared with time-domain variability indices, frequency-domain variability indices were more suitable for short-term 5-min time series analysis for exploring the inherent but slight change due to the emotion effect.