Towards Location Independent Gesture Recognition with Commodity WiFi Devices

Abstract

Recently, WiFi-based gesture recognition has attracted increasing attention. Due to the sensitivity of WiFi signals to environments, an activity recognition model trained at a specific place can hardly work well for other places. To tackle this challenge, we propose WiHand, a location independent gesture recognition system based on commodity WiFi devices. Leveraging the low rank and sparse decomposition, WiHand separates gesture signal from background information, thus making it resilient to location variation. Extensive evaluations showed that WiHand can achieve an average accuracy of 93% for various locations. In addition, WiHand works well under through the wall scenario.

Wi-Fi-Based Location-Independent Human Activity Recognition with Attention Mechanism Enhanced Method

“…This is the key to the realization of human activity recognition. Inspired by [26], to further demonstrate the location-dependent challenges of Wi-Fi-based HAR, we leverage the low rank and sparse decomposition (LRSD) algorithm to separate the original signal into the low-rank and the sparse part, which describe the background and the activity, respectively. Figure 4 displays the sparse component of CSI for the identical activity at four different positions.…”

“…Comparison with different recognition approaches. To verify the superiority of the proposed method, we explore the other two typical approaches, which are CNN and Wi-Hand [26]. CNN is a classical feature representation method that is the most commonly used in wireless sensing.…”

“…Ref. [26] separates the location-related background information from the activity signal, which aims to decrease the number of training locations and samples. Nevertheless, the recognition accuracy is modest and still needs to be improved.…”

Wi-Fi-Based Location-Independent Human Activity Recognition via Meta Learning

“…Some solutions have been proposed to solve the above problems, and remarkable progress has been made, which lays a solid foundation for realizing location-independent sensing with good generalization ability. The solutions fall into the following four categories: (1) Generate virtual data samples for each location [24], (2) Separate the activity signal from the background [25,26], (3) Extract domain-independent features [27], and (4) Domain adaptation and transfer learning. Some approaches involving other domains (such as environment, orientation, and person) can also be grouped into these four categories.…”

“…FALAR [25] benefits from its development of a new Open-Wrt firmware which can get fine-grained Channel State Information (CSI) of all the 114 subcarriers, improving data resolution. Similarly, high transmission rates of the perception signal (such as 2500 packets/s in Lu et al [26]) can also boost the resolution. As the author described by Zhou et al [30], a low sampling rate may miss some key information, which accounts for the deterioration in the system performance.…”