The invention relates to the field of identification of sleep EGG (electroencephalogram) detection signals and discloses a method for classifying sleep stages on basis of sleep EGG (electroencephalogram) signal detection. The method comprises steps as follows: (1), signal acquisition equipment is connected with a left frontal pole and a right frontal pole and is used for acquiring EGG signals, electrooculogram signals and lower jaw electromyographic signals; (2), filtering is performed, the signals are subjected to filtering processing through a low pass filter, and the signals with frequency higher than 50 Hz are cut off, and the signals with frequency lower than 50 Hz are allowed to pass; (3), the signals are subjected to frequency band division, and a time domain is subjected to band division processing with 30s as a unit; (4), characteristic values are extracted, energy of each frequency band is calculated, and an energy ratio is taken as a characteristic value of each frequency band. According to the method, a polysomnography is simplified through the acquisition equipment, then the acquired signals are processed, and the signals of the frequency bands are separated, so that which sleep stage a detected person is in is determined; the overall process is simple to operate, the analysis is automated, and the problem of insufficiency of equipment in a hospital can be effectively solved.