Sleep Movement Detection using k-Nearest Neighbors and Dynamic Time Warping
Analysis of sleep quality and sleep monitoring can serve as a diagnostic feature to check various sleep and psychiatric disorders. The quality of life of a large number of people is affected by diseases like sleep apnea, insomnia, narcolepsy etc. We propose a sleep movement detection algorithm that works accurately to classify sleep movement and non-movement data obtained from unobstructed pressure sensitive textile sheet called Dozee1that uses the Ballistocardiogram (BCG) technique to record signals. The data was collected from 8 subjects, each subject contributing 6 hours of sleep data. We utilized the K-nearest neighbors classification algorithm which is a non-parametric, classical supervised Machine learning algorithm, along with Dynamic time warping method to classify the time-series data and deduce inference from it. The proposed sleep movement detection algorithm achieved an overall accuracy of 98.17%.
Keywords - Machine learning, Time series classification, Sleep movement detection, Dynamic time warping