Features of Smartphone Diagnostics of Human Health Based on Gait Parameters

The article describes trends in the implementation of portable and wearable devices in the field of personalized medicine. The main stages of processing smartphone data to estimate human gait parameters are described. The approaches used are considered and the problems of practical implementation of technology for assessing health parameters using gait parameters using a mobile phone are shown. Special attention is paid to the quality of data collected from the smartphone. To improve the quality of data collected from a smartphone, an algorithm for preliminary automated processing of data from wearable devices is proposed using the example of a smartphone accelerometer. The proposed algorithm takes into account specific time-frequency features of the accelerometer signal in the segmentation procedure and allows selecting the most correlated segments. In practice, the proposed algorithm reduces the appearance of negative anomalous data at the classification stage and increases the reliability of further classification of segments in accordance with changes in human health and increases the reliability of forecast estimates. At the segmentation stage, the probability of selecting false segments is no more than 0.2, and the probability of missing segments is no more than 0.11. Despite the results obtained, the authors note the need to use machine learning methods at the data segmentation stage.