Gait analysis is an important research in the fields of pattern recognition, data mining, and intelligent data analysis. Many different methods, such as recognition of crests and troughs, matching of gait templates, signal-processing based approaches, etc., have been proposed and applied, but these methods require presetting parameters such as gait numbers, gait templates, etc. Their feasibilities are limited. In this study, a new high feasible method, which can automatically calculate a presetting cycle value and a self-adaption range, is proposed, based on peak detection and threshold, to analyze gait cycle segment without unnecessary information needed by other methods. Moreover, a method to filter nonrelated data is proposed. Comparing with simple FFT and other three kinds of representative methods, the proposed approach has the best analysis results in evaluations.