Human motion open-set identification method based on radar image and generation countermeasure model

Abstract

The invention relates to the technical field of radars and the field of human motion, and provides a human motion open-set identification method based on a radar image and a generation countermeasuremodel. The method includes directly distinguishing the input image into a known or unknown motion category and outputting the category information of the input image so as to realize the end-to-end open-set identification of the human motion. Therefore, according to the human motion open-set identification method based on the radar image and the generation countermeasure model, the characteristicof the human micro-motion can be reflected by means of the micro-doppler image of the radar, and meanwhile, a discriminator in the generation countermeasure model is used as an open-set identifier, the input image can be directly distinguished into a known or unknown motion category and the category information of the input image can be output, so that the end-to-end open-set identification of thehuman motion can be realized. The method is mainly applied to the technical field of radars and the human motion identification occasions.

Description

technical field [0001] The invention relates to the field of radar technology and the field of human action recognition, in particular to an open-set action recognition method based on a generative confrontation model. Background technique [0002] In recent decades, human action recognition has attracted extensive attention in many fields. Due to the increasing demand of motion recognition in entertainment, medical monitoring, security, emergency rescue and other application fields, it is considered as a topic with broad application prospects. In the past, human action recognition mainly relied on visual sensor data and achieved many results with the help of computer vision. Afterwards, research in this area was further advanced by depth sensors. Sensors such as the "Kinect" provide researchers with an easy way to obtain depth information. However, these sensors are extremely susceptible to environmental factors such as illumination, occlusion, and weather, and these sit...

AI Technical Summary

Problems solved by technology

The open set recognition method using deep learning has also become an emerging direction. Abhijit Bendale and Terrance Boult proposed a new layer structure - "OpenMax" layer, but this method requires the help of a pre-trained model, so it is in other recognition tasks. usability is not strong

[0006] In summary, existing open-set methods rely on a reasonable choice of probability thresholds, so these methods lack robustness in other tasks

In addition, considering the shortcomings of the above sensors and the advantages of radar, the problem of open-set recognition of human actions using radar micro-Doppler images needs to be solved urgently.

Images