Indoor crowd counting method based on channel state information

Abstract

The invention discloses an indoor crowd counting method based on channel state information, which comprises the following steps of: filtering and smoothing original CSI (Channel State Information) signal data, minimizing noise influence, reducing data complexity through PCA (Principal Component Analysis) dimensionality reduction, converting a wireless time sequence signal into a heat map image through a sliding window, and counting the crowd in the heat map image. And a convolutional neural network (CNN) is designed to perform feature extraction on the heat map so as to map different numbers of people in corresponding environments. According to the method, the problems of high deployment cost and serious sight shielding of a traditional visual image-based method are solved, and the problem of privacy disputes is avoided. According to the method, prediction is carried out by using the CNN network based on the image after CSI conversion, on one hand, signal features are highlighted, on the other hand, correlation between the same category is better obtained, and the performance of indoor crowd counting can be effectively improved.

Description

technical field [0001] The present invention relates to the field of wireless sensing technology and signal processing technology, as well as the field of deep learning, and in particular to an indoor crowd counting method based on channel state information, which is suitable for both static environments with relatively fixed personnel positions and dynamic environments with frequent movement. Background technique [0002] Crowd counting is one of the important technologies for human activity perception applications. Therefore, in many application scenarios, estimating the number of people indoors is of great value for indoor space public safety management and building energy-saving optimization. Shopping malls can also recommend products and Flow regulation. [0003] In the field of computer vision, crowd counting mainly uses image analysis technology, which can be divided into two methods based on detection frame or density map regression. The method based on the detectio...

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the deficiencies of the prior art, the present invention provides an indoor crowd counting method based on channel state information, which preprocesses the original CSI signal data to minimize noise and reduce data complexity, and then the wireless time series signal Converted to a heat map image, and designed a convolutional neural network (CNN) to extract features from the heat map to map different numbers of people in the corresponding environment; problems, and avoid privacy disputes

Images