WiFine: Real-Time Gesture Recognition Using Wi-Fi with Edge Intelligence

Abstract

Gesture detection based on RF signals has gained increasing popularity in recent years due to several benefits it has brought such as eliminating the need to carry additional devices and providing better privacy. In traditional methods, significant breakthroughs have been made to improve recognition accuracy and scene robustness, but the limited computing power of edge devices (the first-level equipment to receive signals) and the requirement of fast response for detection have not been adequately addressed. In this paper, we propose a lightweight Wi-Fi gesture recognition system, referred to as WiFine, which is designed and implemented for deployment on low-end edge devices without the use of any additional high-performance services in the process. Towards these goals, we first design algorithms for phase difference selection and amplitude enhancement, respectively, to tackle the problem of data drift caused by user change. Then, we design a cross-dimension fusion method to extract features of finer granularity from information of different dimensions, thus solving the precision problem of feature granularity. Finally, we design a lightweight neural network architecture by leveraging redundancy to reduce computational cost while ensuring satisfactory recognition accuracy. Extensive experimental results show that the proposed system achieves fast recognition of various actions with an accuracy up to 96.03% in 0.19 seconds.

RF-Alphabet: Cross Domain Alphabet Recognition System Based on RFID Differential Threshold Similarity Calculation Model

“…Finally, we compared the latest gesture recognition algorithms, the comparison results are shown in Table 2 . The RF-alphabet has higher accuracy than RF-FreeGR [ 29 ] and FingerPass in different environments and with different users, and RF-alphabet uses fewer tags to achieve better accuracy compared to RF-FreeGR, although FingerPass [ 25 ] uses WIFI instead of RFID, thus reducing hardware dependency, the model uses a deep learning-based algorithm, and FingerPass requires more training data and is less efficient compared to the model-based RF-alphabet.…”

“…Gesture recognition based on wireless technology has received increasing attention from a large number of researchers because wireless devices have the advantages of being non-intrusive, easy to deploy, and highly scalable. Other wireless technologies such as WIFI [ 25 ], RFID, ultrasonic, and radar have been widely used for gesture recognition [ 22 ]. Yang et al proposed the BVP (body-coordinate velocity profile) model for feature capture in independent domains to make full use of the channel information of WiFi for human activity sensing [ 26 ].…”

Air-GR: An Over-the-Air Handwritten Character Recognition System Based on Coordinate Correction YOLOv5 Algorithm and LGR-CNN

“…Tang et al [ 14 ] designed a WiFi gesture recognition system based on a parallel LSTM-FCN [ 15 ] neural network, extracted different dimensions of gesture by parallel long short-term memory full convolutional network (LSTM-FCN) model features, and evaluate 50 gestures with good accuracy. Tianzhang Xing et al [ 16 ] designed a lightweight gesture recognition system, WiFine, based on Wi-Fi to achieve fast recognition of various actions. Yongpan Zou et al [ 17 ] proposed GRfid, a novel device-free gesture recognition system based on phase information output by COTS RFID devices, to obtain gesture information in a non-contact, non-infringing manner by using RFID tags and achieve better accuracy on the test set.…”

Construct 3D Hand Skeleton with Commercial WiFi

“…WiFi-based Hand Sensing. Current WiFi-based hand sensing applications are built towards specific applications such as gesture recognition [1,22,36,56], sign language recognition [21,30,46,54], finger tracking [37,43,52,53], and keystroke detection [2]. In contrast, HandFi has the capability to support diverse applications.…”

“…However, due to the low spatial resolution of WiFi signals and the small scale of a hand, directly modeling the radio reflection from a hand is challenging. Therefore, existing WiFi-based solutions rely on WiFi signal patterns caused by hand motions to their corresponding hand gestures [17,21,30,36,46,54], or employ geometric constraints to track variations in signal propagation for hand tracking [33,37,43,52,53]. None of the existing WiFi-based handsensing systems directly models the relationship between the hand skeleton and the reflected signals of interests.…”