Find in Library

This paper describes an algorithm for extracting human gait movements in data obtained from accelerometer sensors of a mobile phone, provided that the mobile phone is used in the usual mode. The algorithm also performs a classification of the selected movements based on a feed-forward neural network. The developed algorithm selects the best areas in the accelerometer data, which reflect individual steps, according to the optimality criterion. For the selected area, the optimality criterion is the maximum value of the correlation coefficient with all other data segments. The selected plots are used as templates. Changing the parameters of patterns over time is necessary to assess changes in the individual rate of the functioning of the musculoskeletal system. Due to the correction of tolerance limits at the segmentation stage, the algorithm adapts to the change in gait speed.