Dynamic gesture recognition using wireless signals with less disturbance

As a nonverbal body language, gestures undoubtedly can play a very significant role when interacting with smart devices. One of the most discrete ways of gesture recognition is through the use of Wi-Fi signals. Recent literatures start to explore the feasibility of utilizing the widely deployed Wi-Fi infrastructure to track human motions and interact with smart devices. In this paper, we develop a gesture recognition system, which adopts off-the-shelf Wi-Fi devices to collect fine-grained wireless Channel State Information (CSI). First, low pass filter is used to eliminate noise, then principal component analysis (PCA) is used to reduce data dimension as well as eliminate noise further. Moving objects may have significant disturbance in the gesture recognition and this may occur frequently in the actual environment; thus, we introduce a disturbance eliminating module and independent component analysis (ICA) is used for disturbance eliminate. The experimental results have shown that our system can keep high accuracy even with effects of moving objects.