Linear AR Models for Description of a Stress-Test Heart Rate in Horses

摘要

The paper deals with an algorithm for identifying parameters of an AR model of the HRV data recorded in horses during stress-test. The developed algorithm estimates positions of poles of the 2nd order AR model representing dominant stationary component of the instantaneous heart rate sequences. The stationary component is obtained by means of Hamming low-pass filter with an impulse response, the length of which was optimized to obtain the filtered data as stationary as possible. It was experimentally found that modules of the poles varied in a range 0.963 to 1.0, the smallest value for the horse with the worst performance, the highest value for the best one. Similarly, the phase of the poles or better the frequency corresponding to the phase varied from 0.01 to 0.03 Hz, where the smallest value characterized the horse with the worst fitness and the highest value for the best one.