The Research Of The Gesture Action Adaptive Identification Algotithm

Nowadays, with the development and application of a variety of sensors, thegesture recognition based on three axis acceleration sensor has become a hot researchfield in pattern recognition, and the three axis acceleration sensor can be placed in thepopular Smartphone, therefore, more efficient identification method plays a veryimportant role in the gesture recognition field for their development. In this paper,because the existing gesture recognition algorithm can only meet in the overall optimalbut individuals can not reach the best condition, according to the collected movementcharacteristics and regular pattern of hand signals, an adaptive gesture recognitionmethod based on system identification has been designed, and effectively solved theproblem of individual optimal and recognition accuracy etc.First of all, three axis acceleration gesture signals are collected by accelerationsensor (AMI602). And then the acceleration signal is sent to a txt file saved to a PCthrough the serial port.Secondly, according to the signal requirement of the design, the accelerationsignal is preprocessed, it mainly includes that hexadecimal signal is calculated intodecimal acceleration values, a smoothing process to remove an interference signal,detecting the effective action of the sliding window of data detection, adjusting thesampling frequency of the interpolation process, and normalization of the size of theentire amplitude normalization process.Finally, the adaptive control theory and system identification method is applied togesture recognition, building models by the recursive algorithm of ARX model,ARMAX model based on system identification, and it is elaborated in four aspects ofmodel predicting, model building, model optimization and model validation.The design applys the MATLAB environment as the simulation platform, theexperimental verification and recognition accuracy comparison to the recognition of0~9gestures, the experimental verification of the design method is feasible and obtains a high recognition rate. In experiments, the average recognition rate of the Arab digitalis89.6%.