A WiFi-based Handwritten Signature Authentication And Movement Prediction System

With the study of action recognition based on WiFi has attracted more and more attention,this thesis proposes a WiFi-based identity authentication system for handwritten signatures(Wi Handwritten),which is convenient.The Wi Handwritten uses Butterworth low-pass filter to denoise the raw CSI data.Then the Wi Handwritten extracts the variance,derivative,energy,skewness coefficients and kurtosis coefficients of handwritten signatures from the filtered CSI data as characteristic values to describe signature movement range,speed,and distance between the fingers and the receiving device.Finally,the characteristic values are putted into the proposed LSTM-based authentication discriminator for training to verify the handwritten signature's genuine or forged.The experimental results show that the accuracy of handwriting signature recognition and verification reaches 93.85% and 91.92%,with high robustness.Moreover,this thesis also proposed a WiFi-based movement prediction system(Wi Predict).In the data cutting phase,we need to detect the start and end of the movement on the WiFi signal.Due to different movements need different thresholds,this thesis proposed the variance of variance,which we can set a fixed threshold to detect the beginning and end of the different movements more accurately.