A wireless signal-based gesture recognition method

Abstract

The invention discloses a gesture recognition method based on a wireless radio frequency signal, which uses one signal transmitter Tx and two signal receivers Rx1 and Rx2, and the connection between the signal transmitter and the signal receiver is not parallel; the signal adopts OFDM modulation; the gesture The identification area is the inner plane of the area enclosed by Tx‑Rx1 and Tx‑Rx2; the signal transmitter transmits radio frequency signals, and the two signal receivers collect CSI data streams respectively; the hand is used as the reference frame to construct features, including dynamic phase vectors DPV, motion rotation variable MRV and fine-grained features; use the change of channel state information CSI in wireless communication to estimate the instantaneous relative direction of the hand in the process of moving, and realize the position and orientation and writing through the hierarchical hierarchical multi-layer recognition mechanism. Size-independent robust, high-accuracy gesture recognition.

Description

technical field [0001] The present invention relates to sensorless sensing technology, in particular to a gesture recognition method based on wireless signals, which can recognize gestures in different initial positions, different writing angles and different writing sizes based on commercial WiFi equipment. Background technique [0002] Human-computer interaction technology is an important way of communication between people and execution terminals. People can issue commands through the human-computer interaction interface to drive computers or other hardware execution devices to complete, so as to realize the exchange of information between people and machines and complete specific tasks. Task. Gesture-based human-computer interaction enables natural and efficient information exchange between humans and machines through fast and accurate gesture recognition, providing strong support for many applications. [0003] The existing gesture recognition technologies can be divid...

AI Technical Summary

Problems solved by technology

[0004] Most of the existing gesture recognition methods based on radio frequency signals construct the physical or statistical characteristics of radio frequency signals. However, these characteristics usually take the signal transceiver device as the frame of reference. When the signal transceiver device changes, the signal fluctuation pattern caused by the same gesture will also change accordingly, that is, the consistency between a specific gesture and its corresponding signal pattern cannot be guaranteed, resulting in a great impact on the recognition rate. hinders the application of the method in real scenarios

In order to maintain the consistency of the gesture and the corresponding signal pattern, the method described in the literature (Position and orientation agnostic gesture recognition using WiFi, MobiSys'17, 252-264; Zero-effort cross-domain gesture recognition with Wi-Fi, MobiSys'19, 313-325) Gesture recognition in different positions and orientations can be achieved by building new features or modeling mechanisms, but faced with the need to build machine learning classifiers, increase signal transceiver equipment, etc.

[0005] In summary, most of the existing non-contact gesture recognition methods based on radio frequency signals cannot accurately recognize different gesture starting positions, writing angles and writing sizes, and a few improved methods may require the construction of special machine learning classifiers. Or by adding the number of transceiver devices to achieve gesture recognition with different starting positions, writing angles and writing sizes, all of which limit the practicability and universality of gesture recognition technology

Images