Deep convolutional adversarial neural network-based human body action radar image classification method

Abstract

The invention relates to a deep convolutional adversarial neural network-based human body action radar image classification method. The method comprises the steps of constructing a data set; realizingradar image data enhancement through a DCGAN: establishing the DCGAN, using the network to individually learn each radar spectrogram, generating new radar spectrograms according to characteristics learned by the network, expanding training set samples under the condition of a certain data amount, adjusting parameters by the network to ensure that fewest failure images are generated, and expandingthe data set to the maximum extent, thereby realizing the data enhancement; and extracting upper, middle and lower envelopes in each radar image to serve as eigenvectors, taking the three eigenvectors as inputs of a support vector machine classifier, and classifying radar image data by utilizing a support vector machine, wherein the upper and lower envelopes represent echo radial velocities of human limbs, the middle envelope represents an echo radial velocity of a human trunk.

Description

technical field [0001] The invention belongs to the fields of human action recognition, radar target detection, data enhancement, deep convolutional generative adversarial networks (DCGAN, Deep Convolutional Generative Adversarial Networks) and machine learning, and relates to the feature extraction of radar images and the use of DCGAN for data enhancement and The problem of human action classification. Background technique [0002] Human action recognition [1] is a research hotspot in the field of computer vision in recent years, and it is widely used in the fields of human-computer interaction, virtual reality, and video surveillance. , but the high complexity and variability of human motion make the recognition efficiency and accuracy unable to fully meet the relevant requirements of various industries. The difficulty of human action behavior recognition mainly lies in the complexity of space and the difference of time. Space complexity includes action scenes with diffe...

AI Technical Summary

Problems solved by technology

These two problems make the research field of human action behavior recognition lack a unified and effective framework, related technologies and unified and effective analysis and classification methods.

Images