Respiratory analysis during sleep using a chest-worn accelerometer: A machine learning approach