Outlier detection in heart rate signal using activity information

During exercise, heart rate will increase and its distribution will be different from that in stationary statement. Moreover, activity introduces several outliers into heart rate series. Heart rate variability analysis under exercise conditions can't be conducted identically to the traditional methods. A heart rate distribution analysis method is proposed to fuse the activity information and heart rate signals which will be used to make heart rate analysis under exercise conditions. Firstly we use Gaussian function to fit RR intervals under various intensity activities, and then establish dynamic model for the parameter μ and σ which is changed with activity intensity. With the distribution of heart rate determined, outliers in RR interval can be detected and replaced according to possibility distribution. In the last, the validity of outlier detection algorithm and the influence of outliers to HRV are verified.