A Joint Classifier for Sleep Staging with Pulse Rate Variability Based on Automatic Weight Assignment

Sleep staging is the basis for assessing sleep quality and studying sleep-related diseases. However, standard method to the identification of sleep stages based on polysomnography (PSG) is very inconvenient and may affect the subject's sleep state. This paper proposes a sleep staging method based on a joint classifier using pulse rate variability (PRV). By the pulse oximeter, we obtained pulse to pulse interval series (PP interval series) from the original photoplethysmography (PPG). Then the PP interval series were deeply investigated in time domain, frequency domain and no-linear domain, and 12 features of pulse rate variability were extracted. Finally, by combining the Fisher linear discriminant model, the support vector machine model, and the BP neural network model using automatic weight assignment strategy, we constructed a joint classifier to identify different sleep stages. The feasibility and effectiveness of the proposed sleep staging method has been evaluated by experimental results, which show that the approach can achieve a mean accuracy of 82.19%.