Research On Direction Sensitive Multi Gesture Recognition Method Based On Commercial WiFi Device

As an important part of human-computer interaction,gesture recognition is widely used in smart home and virtual reality.The gesture recognition method based on wearable sensor has the problems of expensive equipment and privacy exposure.Therefore,gesture recognition technology which does not need to wear any hardware devices,does not need to consider the lighting conditions and is low-cost has attracted much attention.This paper focuses on the direction sensitive gesture recognition method.The main work can be summarized as follows.Firstly,a noise elimination method based on the combination of Hampel filter and moving average filter is proposed to detect and remove outliers from the CSI data at the receiving end,and the results are de-noising.Secondly,an automatic gesture detection and dynamic segmentation method is proposed.In order to ensure the accuracy and diversity of gesture samples,firstly,gesture detection is completed by differential dynamic threshold,and then the start and end points of detected gestures are adjusted by gesture dynamic scaling correction algorithm.In addition,according to the principle of Fresnel zone model,the motion characteristics of gesture samples are analyzed.Thirdly,an illegal data detection algorithm based on SVM is proposed to filter the abnormal gesture information.The CSI data determined as legal data is input to the deep neural network for feature extraction and classification.Finally,the full experiment is carried out in the indoor environment,and the experimental results are analyzed.The results show that the average recognition accuracy of the proposed method can reach 84% for the same gesture in multiple directions.