WiGAN: A WiFi Based Gesture Recognition System with GANs

Abstract

In recent years, a series of research experiments have been conducted on WiFi-based gesture recognition. However, current recognition systems are still facing the challenge of small samples and environmental dependence. To deal with the problem of performance degradation caused by these factors, we propose a WiFi-based gesture recognition system, WiGAN, which uses Generative Adversarial Network (GAN) to extract and generate gesture features. With GAN, WiGAN expands the data capacity to reduce time cost and increase sample diversity. The proposed system extracts and fuses multiple convolutional layer feature maps as gesture features before gesture recognition. After fusing features, Support Vector Machine (SVM) is exploited for human activity classification because of its accuracy and convenience. The key insight of WiGAN is to generate samples and merge multi-grained feature maps in our designed GAN, which not only enhances the data but also allows the neural network to select different grained features for gesture recognition. According to the result of experiments conducted on two existing datasets, the average recognition accuracy of WiGAN reaches 98% and 95.6%, respectively, outperforming the existing system. Moreover, the recognition accuracy under different experimental environments and different users shows the robustness of WiGAN.

A Domain-Independent Generative Adversarial Network for Activity Recognition Using WiFi CSI Data

“…Another paper based on GAN is WiGAN [ 7 ], in which the authors showed that their proposed modelis environmental and user independent. First, the authors combined the structure of Deep Convolutional GAN [ 20 ] with the main characteristics of Conditional GAN [ 21 ].…”

“…Furthermore, the authors of [ 7 ] proposed to add a data pre-processing module for the training deep learning model, whose main purpose was to convert raw CSI data into the sanitized CSI amplitude. This module consists of several data pre-processing steps, such as activity detection, interpolation, Discrete Wavelet Transform and sub-carrier selection.…”

“…Data pre-processing plays an important role in addressing the domain change problem in learning-based activity recognition systems using WiFi CSI data. Based on recent research papers, the modern deep learning approaches discussed in Section 3 , proposed data pre-processing pipelines that increase recognition system performance by approximately 20% [ 7 ] or 40% [ 5 ]. However, as discussed in Section 3 , commonly used pipelines in the literature consist of a long list of different methods, and it is quite complex to find an optimal combination of these techniques.…”

“…WiFi CSI contains carrier signal amplitude and phase. Recently, it has been commonly used for fine-grained activity recognition in combination with data-driven, learning-based models [ 2 , 3 , 4 , 5 , 6 , 7 , 8 ].…”

“…However, research has shown that, the performance of these learning-based approaches in the context of WiFi CSI-based activity recognition significantly degrades due to various domain factor changes. For instance, it was shown that performance of WiFi-based activity recognition systems are impacted by a change of the environment in which the activity is performed [ 5 , 7 , 9 ], change of orientation with respect to the sensing device [ 5 ], CSI data quality [ 1 ], different physical properties of human subjects or slight difference in movement patterns of a user [ 4 , 6 ]. Even the time of day may have a big impact, as electromagnetic waves are impacted differently during the day and night in office or home environments [ 4 , 10 ].…”