FingerFi: An Alpha-numeric Character-based Gesture Recognition using Wi-Fi Sensing

Gesture recognition based on WiFi channel state information (CSI) has attracted much attention due to its applications in home automation, robotics, healthcare, and so on. Recognition of alphanumeric characters drawn by a finger in a WiFi sensing zone has its own applications, such as nonverbal communication, accessibility, document digitization, and many more. In this paper, we propose a novel model for alphanumeric gesture recognition using CSI-ratio, the variance of the CSI values in the sub-carriers, and different machine learning models like K-nearest neighbors (KNN), Linear Discriminant Analysis (LDA), Decision Tree (DT). Exhaustive experiments are conducted involving different persons to draw alphanumeric characters using fingers and the CSI values are collected using an Intel 5300n network interface card as a receiver and a TP-Link commodity commercially off-the-shelf (COTS) router as a transmitter. The collected values are pre-processed using the CSI ratio method and the variance of the CSI matrix is used as a feature to classify the alphanumeric gestures using different machine learning models like KNN, LDA, and DT. The DT machine learning method outperforms the other two models and the accuracy of recognizing only digits with DT is 98.26% and 96.24% for both alphabets and digits. We also tried to identify the participating person based on the pattern of their gesture and the DT method is able to detect the participant with 99% accuracy.