Abstract: Sleep scoring is an important research direction in medical research and clinical medicine. Traditional visual scoring method is based on scoring rules, which is a time consuming and subjective procedure. Therefore an automatic sleep staging method based on refined composite multi-scale entropy (CMSE) and multi-level support vector machine is proposed. Firstly, to ensure the reliability of the input characteristics, refined CMSE is extracted as the feature input and two channels of electroencephalogram (EEG) and electrooculogram (EOG) are used. Then a three-layer support vector machine classification scheme is applied to classify sleep stages. Specifically, the inputs of each layer are obtained according to the trend of the entropy curves. The overall accuracy of the proposed method is 85.3%. Compared with traditional methods, the classification accuracy of the proposed method is more balanced and the global performance is much better.

Key words: sleep scoring, refined, composite multi-scale entropy (CMSE)