An Innovative Study on Automatic Sleep Stage Classification Using Feature Extraction and Machine Learning Based on Radio Signal Analysis

With the rapid advancement of radio technology and its widespread application in medical monitoring, this paper proposes an innovative approach for sleep stage classification. The proposed method combines feature extraction techniques with machine learning algorithms, leveraging radio signal analysis to extract feature information from electroencephalogram (EEG) data. The extracted features include power spectral density (PSD), singular value decomposition entropy (SVD Entropy), Higuchi fractal dimension (HFD), and permutation entropy (PE). These features are then integrated with machine learning models such as XGBoost to achieve accurate sleep stage classification. It is demonstrated experimentally that an automatic sleep stage classification method combining multi-feature extraction techniques (e.g., PSD, SVD Entropy, HFD, PE, etc.) with machine learning models (e.g., XGBoost) can significantly improve the accuracy and efficiency of classification. The research presented in this paper not only extends the application of radio signal analysis in the medical field but also provides new insights and methodologies for sleep science research.