In this paper, we demonstrate that WiFi can be used for fingergesture detection which is both effectively deployable and low in cost. Oursystem uses a single WiFi device which is connected to a Access Point (AP) to recognizegestures present in the Channel State Information (CSI). Recognizing correctgestures can get challenging due to environmental changes. For example, aperson walking in the room or a moved chair or furniture can alter the fingergesture in the CSI.
Moreover, individual diversity can also cause distortionsin the finger gesture since different users have different speed of movementand different finger dimensions. Also, same user can also produce variablefinger gesture due to lack of consistency. To deal with environmental noise, we incorporateenvironmental noise removal system which uses wavelet based denoising 1 to reduceenvironmental noises and maintain the CSI gestures generated only from thefinger gestures. In particular, the environmental noise removal systemdecomposes the CSI into details and approximations and removes any unwantednoise above a predetermined threshold by thresholding.
This helps to maintainthe necessary details while neglecting any unwanted noise. To deal withindividual diversity, our system identifies principal components in the CSIpattern. By extracting the principal components, the system identifies thoseparts of the CSI which are invariant across multiple instances and this makesthe system immune to individual diversity and gesture inconsistency.