Gesture recognition method based on WiFi signal

Abstract

The invention discloses a gesture recognition method based on a WiFi signal. The gesture recognition method comprises the following steps: S1, collecting CSI data of 0-9 number gestures by using a notebook computer and a router which are provided with an Inter5300 wireless network card; S2, extracting an Ntx*Nrx*30 three-dimensional CSI matrix from the collected gesture data; S3, performing data preprocessing on the extracted CSI matrix data; s4, performing feature extraction and reconstruction on the pre-processed CSI matrix to construct a matrix which can be processed by a convolutional auto-encoder; and S5, establishing a convolutional auto-encoder model, and performing feature extraction and classification on the CSI matrix obtained in the step S4. According to the method, a user does not need to wear or depend on any sensor and only needs to use ubiquitous WiFi to process channel state information in the WiFi and then use a convolutional auto-encoder to perform feature extraction and classification on the channel state information, so that identification of ten gestures is realized.

Description

technical field [0001] The invention relates to the technical fields of wireless communication and artificial intelligence recognition, in particular to a gesture recognition method based on WiFi signals. Background technique [0002] With the development of wireless communication and artificial intelligence technology, gesture recognition is becoming an important pillar of smart home automation and human-computer interaction, and gesture recognition has great potential for many other emerging applications in home automation, such as adjusting temperature and brightness levels , for personalized thermal comfort, and remote control of household appliances. Some studies have proposed the use of wearable devices to identify human body activities. However, this method has not been widely promoted because wearable devices are expensive and inconvenient to carry. Some studies have proposed the use of RGB cameras and infrared depth positioning cameras for gesture recognition based...

AI Technical Summary

Problems solved by technology

Some studies have proposed using wearable devices to identify human body activities, but this method has not been widely promoted because wearable devices are expensive and inconvenient to carry.

Some studies have proposed the use of RGB cameras and infrared depth positioning cameras for gesture recognition based on vision systems. Although their recognition accuracy is satisfactory, the need for favorable lighting conditions and potential privacy issues hinder their popular implementation.

There is also the use of WiFi channel state information CSI to realize gesture recognition, but most of them use traditional machine learning to build recognition systems. The data denoising and training process is cumbersome and requires professional knowledge. The gesture classifier uses KNN and other algorithms. To achieve a certain recognition effect, but the traditional machine learning technology needs human operation when extracting representative features related to various gestures, which will lead to a decrease in recognition accuracy

Images