A Fine-grained volley gesture recognition method with Direction Independence

The development of wireless sensing technology makes it possible to control various IoT devices with gestures. In the current severe international epidemic situation, the use of non-contact gesture recognition methods can greatly reduce direct or indirect contact between people and effectively curb the spread of the virus. However, since the direction, position, and motion range of each person's gesture are different, it brings great challenges to the accurate recognition of gestures. To solve this problem, this paper proposes a direction-independent fine-grained volley gesture recognition method CSI-VGR based on Wi-Fi signal sensing technology. CSI-VGR first converts the CSI signal of a gesture into a Doppler shift. In particular, CSI-VGR uses a temporal convolutional network (TCN) to segment the gesture action and extract the features of each gesture. Finally, the Echo State Network (ESN) is introduced, and the gesture feature is input to obtain the classification result of the gesture. This paper verifies the performance of CSI-VGR in multiple dimensions in two real scenarios. Experimental results show that the gesture recognition accuracy of the CSI-VGR method can reach 96.3%, and it shows satisfactory robustness in both scenarios.