With the development of medicine and the improvement of people’s living standards, the issue of rehabilitation is getting more and more attention. Gait rehabilitation provides a brand-new treatment method for patients with walking disfunction. It is currently recognized as an advanced rehabilitation medical method in the world. In recent years, the number of patients suffering from dyskinesias in the lower limbs in China has been increasing, and the society's demand for walking rehabilitation treatment is also increasing. The emergence of gait rehabilitation solves the problem of fewer therapists and more patients, reduces the intensity of the therapist’s work, and has the incomparable advantage that traditional rehabilitation methods lack. However, because there are no mature related products in China at present, and the prices of foreign products are very expensive, domestic medical institutions have not yet put them into practice. Accelerating the development of gait re-habilitation equipment is of great significance for improving China’s medical level, improving the quality of life of patients, and reducing social burden. Usually, high-precision optical sensors are installed on human limbs or using high-speed cameras to capture motion. However, due to the high cost of the equipment, the relative high price of image processing software when processing the collected motion data. In this paper, the acceleration sensors are installed on the human body and the data in a gait cycle can be obtained. After smoothing, using it as the input signal for gait feature extraction and classification. In order to classify normal gait and abnormal gait for evaluation and better walking rehabilitation.