The Research On Sleepiness Detection Based On ECG And Pulse Signals

With the high pace of modern life, lack of sleep and difficulty in sleeping, eveninsomnia often happen. In the pursuit of the high efficiency life, meanwhile, we alsosuffer from fatigue and sleepiness. In high-risk operating environment, such asdriving, steel area and nuclear industry, if operators were sleepiness in working,people’s life and property safety will be threatened. And in the condition monitoringof some patients, such as obstructive sleep apnea syndrome, real-time monitoring ofsleepiness is very necessary. Electrocardiosignal (ECG) and pulse signals are rich inhuman physiology information, and they are convenient in measuring. So, in thisstudy, sleepiness detection experiments were designed, and ECG and pulse signals of30healthy subjects without sleep problems were synchronizing collected.After the preprocessing of the ECG and pulse signals which were selected from26subjects, through feature extraction and analysis, some features were selectedwhich had significant difference between awakening and sleepiness states. Insleepiness state, the amplitude of T wave in ECG, the height of dicrotic wave andpulse transit time very significantly reduced (p<0.005), the time from pulse startingpoint to the peak point of main wave in pulse very significantly increased (p<0.005),heart rate value and VLF of heart rate variability significantly reduced (p<0.05),comparing with the awakening state. Linear discriminant analysis and support vectormachines were used to feature classification of these single features or combinedfeatures.This paper extracted the wavelet energy spectrum of ECG signal and more than20kinds of ECG and pulse features in time domain and frequency domain. Besides,the magnitude values of RR period and pulse characteristics K value were analyzedin continuous time on the sleep-deprivated subjects’ signals.The results of this study show that: the sleepiness has influence in some ECGand pulse features; the amplitude of T wave in ECG, the height of dicrotic wave,pulse transit time, the time from pulse starting point to the peak point of main wavein pulse, heart rate and VLF of heart rate variability have significant differencebetween awakening and sleepiness states; feature combination is helpful to improvethe classification accuracy rate; combined features which the amplitude of T wave inECG takes participation in are higher than the others in classification accuracy rate.ECG and pulse features can be used to distinguish awakening and sleepiness states, considering the convenience of ECG and pulse signals measurement, it is expectedas the new method of sleepiness recognition.