The invention relates to an automatic sleep staging method based on a multi-parameter feature combination. The method includes the steps of collecting EEG signals, EMG signals, ECG signals and respiration signals, denoising all signals, extracting energy ratios of alpha, beta, theta and delta characteristic waves of the EEG signals, extracting the sample entropy of the EEG signals by a sample entropy algorithm, extracting the high frequency characteristic energy ratio of the EMG signals by a wavelet decomposition algorithm, extracting the sample entropy of the ECG signals by the sample entropy algorithm, extracting the mean value of the respiration signals by an averaging method, inputting the five feature parameters into a support vector machine for training and testing, thereby obtaining classification results. According to the automatic sleep staging method, the method of extracting EEG, EMG, ECG and respiration multiple characteristics is adopted to greatly improve the accuracy and generalization ability of sleep staging. The experimental results are reliable and accurate in sleep staging, thereby providing an effective basis for assessing sleep quality and being of a good application prospect.