Human Behavior Recognition Based on WiFi Channel State Information

Nowadays, with the continuous development of WiFi technology, more researchers come to realize that human behavior can be recognized by the application of WiFi Channel State Information (CSI). When human behavior has some changes, it will influence reflections of WiFi signals, which will also cause some changes to the CSI. Using the Intel WiFi Link 5300 network interface controller (NIC) and CSI-Tool, we can obtain the CSI data of corresponding behaviors. In this paper, we design a system to recognize different human behaviors based on the Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU). Firstly, we use the original data collected by the CSI-Tool, then extract the CSI amplitude values of different behaviors as features and input them into neural network structures where the GRU and CNN are connected in parallel. Based on the above works, we can successfully identify different human behaviors.

I. Introduction

Human behavior recognition mainly includes wearable sensor-based behavior recognition [1], vision-based behavior recognition [2], and radio frequency-based behavior recognition [3]. Among them, although the wearable sensor can realize fine-grained motion recognition with high accuracy by collecting the information on environmental changes, the system still has some severe problems to solve. The first problem is the portability of the system. Most of the systems require users to wear individual sensor devices, which make users feel inconvenient and restrict their freedom. Besides, they also have operability issues. Sometimes, they require multiple sensors to implement functions, while users can not complete the use of any time and position. Above all, to get a better user experience, vision-based behavior recognition systems come into being. These systems mostly use cameras to collect video or image information of behaviors, and then use image processing and computer vision technologies to identify behaviors. However, this method has the issue of privacy. Not everyone is willing to be remembered by the camera. Considering the problems of the above two technologies, the behavior recognition system based on WiFi comes into being.