In pattern recognition of golf swing, it is most important to catch the timing (the start, the peak and the finish, etc.) accurately. Therefore, in this study, propose a method that can identify golf swing with high accuracy by using Hidden Markov Model (HMM), which is often applied to the field of time series analysis of statistics, for example, voice recognition, motion recognition. In this paper, compare normal distribution model that has been used in previous study to HMM.