Single-channel EEG automatic sleep staging based on transition optimized HMM

Sleep staging is a key process for evaluating sleep quality and diagnosing somnipathy-related diseases. Psychologists are required to do the traditional sleep stages identification. The manual work by these experts is often time-consuming and error-prone. In order to improve the performance of such a process, automatic Electroencephalography (EEG) signal analysis using machine learning approaches is often used. In this paper, a transitions-optimized Hidden Markov Model (HMM) model is proposed to improve the accuracy of prediction. Our proposed framework includes 4 key modules: feature extraction, feature selection, classification, and transition optimization. By applying, transition optimization process after a general GMM-HMM classification, our experimental results show a convincing improvement in the accuracy of classification.