| Handwriting recognition(HWR)is a typical human-computer interaction application,which can obtain the handwritten track written by the user on the device for information entry and signature authentication.Most of the existing handwriting recognition systems use electronic screens or cameras and other expensive hardware devices to achieve,some of which have hidden dangers of privacy disclosure,and some require users to hand-hold or wear devices to write,reducing the convenience of use.With the development of Internet of things technology,handwriting recognition technology has become a research focus.And the application field of handwriting recognition has also been broadened.Radio Frequency Identification(RFID)technology provides a possibility to realize a handwriting recognition scheme with stronger interaction and higher convenience.Therefore,this paper proposes a device-free handwriting recognition scheme based on RFID tags to solve the existing problems in handwriting recognition.The main research contents are as follows:First of all,in view of the abnormal signal fluctuation caused by ambient noise and equipment defects,the system firstly performs phase unrolling operation to eliminate phase jump and restore disconnected phase.Then the noise is eliminated by gaussian weighted moving average filter and the continuous phase flow associated with handwriting is obtained.Then,a segmentation algorithm based on differential signals is proposed to analyze the characteristics of handwritten signals and realize the segmentation of continuous phase stream.The algorithm obtains the phase fragment corresponding to each character by finding the initial and ending points of the handwritten fragment.Secondly,according to the user’s handwritten trajectory,the system performs finger position positioning and trajectory reconstruction.According to the characteristic that phase can describe the distance relationship,a signal reflection model is established,and the relationship between finger moving position and label phase change is described by analyzing signal propagation path.Then,the continuous handwritten trajectory is decomposed into finger position coordinates corresponding to each sampling point,and the finger position coordinates corresponding to each sampling point are located using the signal reflection model.The finger position coordinates of each sampling point are connected in time sequence to reconstruct the handwritten trajectory.In order to prevent misrecognition of handwritten trajectories caused by rotation or reversal of reconstructed trajectories,trajectory transformation is carried out to unify handwritten trajectories in different directions and positions,so as to achieve accurate handwritten recognition.Finally,commercial RFID devices are used to deploy the system in two scenarios,and the handwriting recognition performance and trajectory tracking performance of the system are evaluated through multiple groups of comparative experiments.The experimental results show that the average recognition accuracy of letters can reach 94.08%,and the tracking error of handwritten track is controlled at about 1cm. |
