Machine Learning-Based Wireless Perceptual Action Recognition Method

Abstract

The invention discloses a machine learning-based wireless perception action recognition method. The invention includes data collection, data denoising processing, feature extraction, SVM model training and recognition processing. During data collection, the absolute value of a set of CSI data collected at each sampling point is taken and read into a matrix form of 30×Nr×Nt. Data denoising mainly adopts PCA method. In feature extraction, it is realized based on discrete wavelet transform. In order to facilitate the training of the SVM model, the present invention performs Kmeans clustering processing on the training samples, uses the obtained n cluster centers as word bags, votes based on the feature vector and the most matching item of each word bag, and uses the feature vector in matrix form When converted to a column vector, it is convenient for the implementation of SVM. The implementation of the present invention provides a human behavior recognition method with high recognition accuracy and good robustness to environmental changes.

Description

technical field [0001] The invention belongs to the field of artificial intelligence technology, and in particular relates to a machine learning-based wireless perception behavior recognition method. Background technique [0002] In today's society, with the continuous development of information technology, as an important research field in artificial intelligence technology, the development of human behavior recognition technology has become a key technology to promote progress in fields such as health care, smart home, and health status tracking. Important position. Traditional human motion recognition technology mainly uses camera equipment, radar or some wearable sensor equipment. However, vision-based methods are limited by user privacy and lighting conditions. Operation with radar is limited by its operating range. However, some wearable sensor devices are not portable enough and inconvenient to use. At the same time, with the development of society and the increas...

AI Technical Summary

Problems solved by technology

However, a quantitative feature is not proposed to describe the relationship between CSI data and different human actions, so that the designed action recognition model shows good robustness in different environments.

Moreover, the traditional low-pass filter and medium-pass filter used in most researches to denoise the collected CSI data, the denoising effect is not particularly ideal, and better denoising technology is needed

Images