Author(s)

Zheng Fufu, Su Yan and Li Dezhao

Corresponding Author

Li Dezhao

ABSTRACT

Sleep plays an important role in human health. A sleep stage classification method for healthy people and patients with Elman neural network was proposed in this work. The classification process included four essential steps: data acquisition, signal preprocessing, feature extraction, and classification. Wavelet threshold denoising and wavelet packet transform were applied for the signal preprocessing. With the Elman network, the accuracy for healthy people is 90.48% and 82.36% for patients, respectively. Besides, the skewness and kurtosis of six characteristic waves were selected with higher relevance to the sleep stages and less redundancy to other features. This study presented a sleep stage classification method with well generalization performance and conclusions, which contribute to the EEG signal analysis for healthy and slight sleep disorders individuals.

KEYWORDS

Sleep stage classification, elman network, EEG signals