Heart Rate and Heart Rate Variability as Classification Features for Mental Fatigue Using Short-Term PPG Signals Via Smartphones Instead of ECG Recordings

The real-time detection and prediction of mental fatigue, mood and stress have received more and more attention the last few years. Mental fatigue can bring health hidden trouble to human body, make the women suffer from mammary gland cyst and uterine fibroid extremely easily, the men suffer from liver cyst and thyroid tumor. Photoplethysmography (PPG) technology is more suitable than Electrocardiography (ECG) for the real-time detection of human physiological signal via smartphones, smartwatches, and wearable sensors to prevent fatigue. Since PPG signal is vulnerable to interference, the polynomial fitting method and Savitzky-Golay (SG) filtering method were used to remove baseline wander and smooth waveform. Then, the adaptive peak-seeking algorithm was used to extract the R-peaks, and the heart rate (HR) were calculated based on R-R intervals (RRIs). The Welch spectrum estimation was used to obtain the spectrum diagram, and high-frequency component power (HF), low-frequency component power (LF) and the ratio of high-frequency component power and low-frequency component power (LF/HF) of heart rate variability (HRV) were calculated. The results of feature analysis showed that RRIs and frequency domain characteristics would change with the degree of mental fatigue rooting in physiological activities in a day. In the afternoon, HR would increase, while LF/HF and LF would decrease. In the evening, when the mind was extremely tired, HR would decrease, while LF/HF and LF would increase. The feature analysis also showed that HR of girls (76.15±10.462) was significantly higher than that of boys (70.82±10.326). These results of feature analysis indicated that sympathetic and parasympathetic nerves would self-regulate according to different mental fatigue levels, resulting in changes of HR, frequency domain characteristics of HRV. Further, the accuracy and specificity of mental fatigue classification were achieved 92.26% and 96.12% respectively. Therefore, we can detect the degree of mental fatigue based on HR and HRV in practice.