Research Of Gesture Recognition Algorithms On Smart Watch

In recent years,with the popularity of wearable devices,smart watches,the emerging products quickly seize the market and continue to be widely concerned by the public.As a symbol of portability and ease of use,smart watches are asked for providing better interaction experience.Hand gesture is simple,efficient and unaffected by external environment,which makes it very suitable to work on smart watches.Currently,the application of hand gesture on smart watches is still quite limited,most of which are based on inherent rule matching.In this paper,we try to put forward a generic recognition solution for mature watch products,and verify its effectiveness by experiment,thus providing a feasibility basis for the function expansion of the smart watches.Firstly,we study current commonly used gesture recognition technology and choose the acceleration-based algorithm as final solution according to the pros and cons.We model the time series data in the light of actual needs,and propose an algorithm framework based on dynamic time warping and hidden Markov model considering the platform constraints.Subsequently,the corresponding auxiliary algorithms are carefully designed taking into account the special characteristics of the gesture data.Among them is a newly raised data interception method that employs longest continuous subsequence and backtracking,which leads to effective completion of data processing.Then,based on the algorithm model proposed above,we implement a gesture recognition system on Ticwear operating system.Finally,a complete experiment is designed according to the research requirement,after which the system mentioned above is adopted to verify and evaluate the hypothesis models.The results show that two models can give the prediction results within the delay of 300 milliseconds,and both achieve an average recognition accuracy over 80%.To sum up,the feasibility of the solution is proved.