Machine Learning Sleep Phase Monitoring using ECG and EMG

Abstract

Sleep is one of the essential parts of living. Lack of sleep may result in concerns and may also indicate underlying health conditions. Hence, the study focuses on determining the sleep phase using data extracted from the Arduino AD8232 (ECG) and Myoware (EMG) sensor to evaluate heart rate variability and EMG Power, respectively. Feature extraction using Machine Learning assisted in interpreting the data acquired from both sensors and comparing results using a commercial-grade smartwatch. The study dealt with several tests to obtain samples from people ages 14–50 years old for at least 2–3 hours to complete a whole sleep cycle. The data extracted were trained using SVM-KNN in MATLAB and Python. The proposed system model resulted in an accuracy of 64.57% for classifying sleep phases and 94 % for sleep and wake.