This paper is structured as follows: Section II provides an overview of radar-based concealed object detection, GANs, and GAN applications for radar data. As such, this paper presents the design a proof of concept for the use of GANs in radar signal generation with a focus on concealed object detection and seeks to establish the foundations for further research into the application of GANs for radar signal generation. In the absence of such data, which is currently the case on current radar based concealed object detection methods, these algorithms fail to exceed the performance of human inspection of radar data, which is labourious and expensive. Applying advanced machine learning algorithms such as deep neural networks Moreover, there have been recent endeavors which analyze the utility GANs for data augmentation. Recent research on GANs have been focused on image generation and, as a result, GANs for one-dimensional data are still in the early stages of development. Generative Adversarial Networks (GAN) have been a popular method of unsupervised learning in computer vision in recent years. The application of machine learning methods to automate radar based concealed object detection using has been limited by the lack of availability in high quality radar signal data.
Generally, these procedures involve computer vision problems which frequently apply machine learning methods to automate the process. There has been a recent trend in this field to use a multimodal screening procedure for deceptive behaviour.
In security and access control systems, concealed object detection plays an integral part of ensuring public safety and security. The application of radar based concealed object detection has been used in areas such as buried landmine detection, buried root detection, breast tumour detection, and concealed weapon detection on people. Radar based methods are commonly used to nondestructively detect concealed objects. Index Terms-generative adversial networks, radar, concealed object detection, deep learning The proposed GAN generated radar signal data which was indistinguishable from the training data by qualitative human observers. Data collected using the Finite-Difference TimeDomain (FDTD) method on three concealed object classes (no object, large object, and small object) were used as training data to train a GAN to generate radar signal samples for each class. As such, this paper proposes the design of a GAN for application in radar signal generation. Generative adversarial networks (GAN) have shown promise in data generation application in the fields of image and audio processing. University of Calgary Biometric Technologies LaboratoryĬalgary, Canada Yanushkevich Department of Electrical and Computer EngineeringĬalgary, Canada major obstacle in radar based methods for concealed object detection on humans and seamless integration into security and access control system is the difficulty in collecting high quality radar signal data. Thomas Truong Department of Electrical and Computer Engineering Generative Adversarial Network for Radar Signal Generation
0 kommentar(er)
