Analyzing Sleep Stages in Home Environment Based on Ballistocardiography

作者：Ni Hongbo; Zhao Tingzhi^*; Zhou Xingshe; Wang Zhu; Chen Lei; Yang Jun

来源：4th International Conference on Health Information Science (HIS), 2015-05-28 to 2015-05-30.

DOI：10.1007/978-3-319-19156-0_7

• Heartbeat interval
• BCG
• Wavelet analysis
• Sleep stage

Currently, a number of people have various sleep disorders, and sleep stages play an important role in assessment of sleep quality and health status. This paper proposes an effective approach of analyzing the sleep stages based on ballistocardiography (BCG), which can be continuously detected with micro-movement sensitive mattress (MSM) in this work, during non-intrusive sleep in home environment. This paper focuses on extracting features from BCG from the following three aspects: multi-resolution wavelet analysis of the heartbeat intervals based time-domain features, Welch's power spectrum estimation based frequency-domain features and the detrended fluctuation analysis (DFA) value for long term correlation based features. Moreover, the support vector machine (SVM) with or without the factor of sleep rhythm, and recurrent neural network (RNN) are adopted to build the classifiers, and both the personal model and self-independent model are investigated for different scenarios. Experimental result of 56 subjects [25 women and 31 men, aged from 16 to 71] was evaluated applying the proposed method and compared to the result provided by professional visual scoring by ECG and EEG. The SVM with the factor of sleep rhythm shows better performance with an average accuracy between 73.21%similar to 83.94% in the personal model, and the self-independent model also achieves a satisfactory level with an average accuracy of 73.611 similar to 78.78% for male and 73.99%similar to 79.46% for female.

• 出版日期2015
• 单位西北工业大学