Calligraphy Writing Motions Recognition Based On Wi-Fi Signal

Chinese calligraphy is a popular and highly esteemed art form in the Chinese culture sphere and worldwide.Writing with correct strokes order is essential to the aesthetics beauty of C hinese calligraphy.However,mastering the correct stroke order is a challenging task for learners,because many C hinese characters do not follow the general rules.In recent years,with the deepening of activity recognition based on Wi-Fi,the fine-grained activity recognition has been continuously enriched.Therefore this thesis presents a novel approach to assist calligraphy learners in getting the stroke order right a kind of calligraphy handwriting motions recognition based on Wi-Fi signals.O ur approach uses the channel state information(CSI)of Wi-Fi signals to device-free track the user’s hand movements during writing,and exploits such information in combination statistical methods and machine learning techniques to infer what C haracter have been written and at which stroke order,then feedback the recognition result to user in time.O ur prototype system is re ferred to as Wi-Calligraphy.This paper is specifically divided into the following two aspects:1)To solve the problem that the writing motions’ CSI data cannot be matched or the method is not applicable after the change o f physical environments or user writing modes.This paper proposes a solution to calibrate the CSI data.We analyze the mapping relationship between strokes and CSI data in different scenarios or user writing modes.Then we use a calibration function to perform deployment conversions for different physical environment and the user’s writing mode.2)To solve the problem that undefined C hinese characters writing motion recognition.We propose the CSI data segmentation and recognition scheme for calligraphy wr iting motions.We only train the basic strokes writing motions data,and the character data is divided into the strokes data,then identify these strokes and recombine these strokes for undefined or untrained characters recognition.We have conducted extensive experiments and user studies to evaluate our approach.Results show that our approach is highly effective in identifying the written characters and their written stroke order.We show that our approach through calibration can adapt to different deployment environments and user mode.