Detection Of Stress-Sate Based On PPG Signal

With the rapid progress of the current society,great changes have taken place in social ideology.People are facing more and more stress from family,society or themselves.Negative stress can cause mental and physical pain.Therefore,it is necessary for people to monitor their stress-state and make adjustments in time.Among the stress-state detection methods for the human body,smart bands are favored by many consumers due to their advantages such as noninvasiveness,simple operation,and small size.The stress-state detection function of smart bands is obtained based on pulse wave signals through the photoelectric sensor,called Photoplethysmography(PPG),which contains rich information such as cardiovascular function including real-time heart rate changes.The existing stress monitoring functions are mainly based on heart rate variability indicators.Heart rate variability is an important indicator for evaluating the function of autonomic nervous system,which quantifies real-time heart rate changes.Not only pulse rate variability(PRV,an alternative indicator for HRV)can be extracted from PPG,but also the morphological features of each cardiac cycle can be extracted,which can also be used for evaluating stress-state.At present,relevant researches on stress-state assessment based on PPG signals mainly focused on PRV,and there were few studies analyzing stress-state through PPG morphological characteristics or combining the two characteristics.The current PPG feature extraction methods were often based on a single database or a specific PPG device,and thus the proposed algorithms had limitations.Therefore,new algorithms were proposed in this thesis to extract PRV indicators and PPG morphological features for multiple PPG-based stress detection databases,and further the stress-states were accessed by machine learning models.The main research contents are as follows:(1)A stress-inducing experiment was designed.The PPG signals were recorded from 30 subjects who were wearing two brand wrist bands at the same time.The experiment procedure included deep breathing,resting state and different pressure states.Further,PPG features were extracted to evaluate stress-state,and the feature extraction algorithms were verified by two public PPG-based stress detection database.(2)Due to the mobility of the wearing position of smart bands,the signal-to-noise ratio of PPG can be easily affected by activities.In order to remove noise and obtain high quality realtime heart rate signals,a new method was proposed to extract PPG real-time heart rate signals based on the personal thereshold of heart rate changes.The results showed that this method had good denoising effect in three databases.(3)The existing methods to extract the characteristic points of special morphological parameters of PPG waves were mainly aimed at the PPG signals with standard waveforms,but the extraction performance for non-standard PPG wave was poor.Therefore,to solve this problem,a new algorithm was proposed to extract the special characteristic points of PPG pulse waves.In addition,a quality evaluation method was proposed for the extracted PPG morphological parameters based on physiological significance of PPG morphological feature points.The effectiveness of the algorithm was verified by calculating the proportion of retained data points of morphological parameters.The results of three databases showed that the proposed algorithm could effectively improve the quality of extracted morphological parameters.(4)The PRV indicators and PPG morphological parameters were applied to the stress-state detection of three types of databases.The results showed that the algorithms proposed in this thesis were effective in three databases.The average accuracy in the database collected in this thesis was 83.87%,and reached 99.79% and 98.83% respectively in the other two public databases.