Machine Learning Based Sleep Phase Monitoring using Pulse Oximeter and Accelerometer

Having enough sleep is essential to health. Tracking a person's sleeping pattern using an assistive device is beneficial in improving their sleeping habits and getting a good quality sleep. The study created a device that monitors the sleep stages using a pulse oximeter and accelerometer. The subjects were put to sleep with the prototype placed on the wrist. Heart rate and acceleration rate signals for 90 to 120 minutes were acquired using the two sensors. Data were collected and analyzed using Fast Fourier Transform for feature extraction. Support Vector Machine was used to classify the sleep stages, namely, Non-Rapid Eye Movement (N1, N2 and N3) and Rapid Eye Movement (REM). Promising results in detecting sleep stages were obtained and validation was performed using a commercial device. The accuracy rate in classifying the N1 stage for all respondents is 91.92%, for N2 stage it is 95.19%, for N3 stage it is 96.06% while for REM stage it is 91.57%.