Heart Rate State Recognition Under Different Behaviors

The detection of human physiological health parameters is a hot research direction in the field of "intelligent medicine".Heart rate is a direct indicator of heart health.Since the heart rate detection of most wearable devices only detects the current heart rate value,it cannot automatically detect the heart rate status under different behaviors.Therefore,in this paper,human physiological monitoring as a starting point,from the two aspects of behavior recognition and heart rate detection,and the heart rate status detection under different behavior software design.The main contents include: improving the wavelet denoising algorithm,putting forward SPLDA-XGB behavior recognition algorithm,and designing the heart rate state recognition scheme under different behaviors.An improved wavelet denoising algorithm was proposed to solve the problem of peak or abrupt noise in acceleration signals.The improved wavelet threshold function and composite evaluation index are used to denoise the resultant acceleration signal.Experimental results show that the improved wavelet denoising algorithm can effectively remove the peak or abrupt noise of acceleration signal and provide effective behavior signal for feature extraction.In view of the "dimension disaster" caused by data redundancy in the process of behavior recognition,SPLDA-XGB behavior recognition classification algorithm is proposed.First based on PCA and LDA algorithm SPLDA dimension reduction algorithm is proposed,SPLDA algorithm can guarantee under the condition of the original sample covariance structure remains the same,the most important principal components in the transformation matrix are obtained for empowerment,by adjusting the divergence between the matrix and the class divergence within the matrix,makes the class while maximizing the distance between the class in the distance minimization,and gain more optimal projection vector,update the class divergence within the matrix,the obtained sample data sets to dimension reduction.The sample data set after dimensionality reduction after SPLDA is classified by XGBoost classification algorithm.The experimental results show that the SPLDA-XGB classification algorithm improves the accuracy and recall rate of behavior recognition,and reduces the recognition time.Aiming at the problem that heart rate status can only be judged by human under different behaviors at present,the design scheme of heart rate status identification software under different behaviors is proposed.By calibrating the three level threshold value of the normal heart rate state range under different behaviors,the heart rate state identification under three behaviors of stillness,walking and running is preliminarily realized.The research work of heart rate state identification under different behavior in this paper can provide important reference value for its application in weara ble devices.