Research On Lip Language Recognition Methods Based On RFID

With the rapid development of wireless sensing technology,lip recognition technology has gradually become a research hotspot because it can recognize the text content conveyed by extracting the lip features of the user when speaking.As RFID technology,which is a non-contact data communication,has been popularized,RFID-based lip recognition methods have been developed.However,the existing research results have the following two challenges:(1)it is difficult to capture the large frequency shifts induced by multiple tags due to stretching;(2)the RFID tags are easy to fall off,and the phase deviation problems caused by the RFID hardware system.Therefore,this thesis proposes an improved DTW lip recognition algorithm based on RFID,the researches are as follows.(1)An RFID-based contact lip recognition method is proposed.In response to the problem of how to extract effective discriminative features from the lip data collected by RFID,this chapter establishes a processing model based on the RSS and phase variation collected by RFID readers.When removing the mutations and noises existing in the original data,the relationship between curve smoothing and curve characteristics should be balanced,the Gaussian function and S-G filter are used to smooth and denoise the noise pre-processing of the original data.Furthermore,the RFID-based normalized DTW lip recognition algorithm is introduced to evaluate and analyze the collected signal features.The problem of pronunciation length mismatch is dealt with while normalizing the data.(2)An RFID-based non-contact lip recognition method is proposed.To deal with the problem of tag shedding and loss in contact lip recognition,research on the phase change of RFID signal for non-contact lip recognition is carried out.The threshold method is used to detect the radix ambiguity and phase entanglement problems introduced by RFID system hardware.The above interference is removed by adding and subtracting pi-values.The polarization and then difference strategy is proposed to segment the time series data set.The DTW distance matrix of the segmented dataset is extracted,and the RFID-based K-Nearest Neighbor DTW lip recognition algorithm is used for classification.(3)The RFID-based lip recognition prototype system is built to accomplish tasks such as the collection and recognition of lip recognition time series data,further verifying the feasibility of the RFID-based lip recognition method.This thesis analyzes and verifies RFID-based contact and non-contact lip recognition methods.Experiments show that both methods can achieve lip recognition.Compared with the traditional lip recognition method,the RFID-based lip recognition method proposed in this thesis has the characteristics of strong applicability,small limitation,and low cost.For recognizing the digital signal characteristics of different users,the method in this thesis has a large improvement compared with similar methods.This thesis has 40 figures,10 tables,and 80 references.