Object classification on raw radar data using convolutional neural networks
Authors:
- 1 March 2019
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Abstract
This paper evaluates the classification of objects given their signal data via a simple convolutional neural network (CNN). Many of the signal processing neural networks involve sound frequency data or Doppler signatures that contain the characteristic features of each object. In this study, we use frequency-intensity data within range-time domain from a Frequency-Modulated Continuous-Wave (FMCW) radar to classify detected objects. The application of various data augmentation methods mitigated the scarcity of labeled data from our field experiments. Time stretching, frequency shifting and noise addition preserved the semantic information of each rangetime data, further improving the models ability to generalize. Modifications applied to our data, which is then converted into a low-level log-scaled mel-spectrogram representation, are learned by CNN models with a set of convolutional and max-pooling layers along with fully-connected layers and selective residual module. Based on our experiments, we conclude that raw radar data can be used for training CNNs for classification and thus can be used to classify a car, a human, and an UAV.Keywords
This publication has 11 references indexed in Scilit:
- A Study and Comparison of Human and Deep Learning Recognition Performance under Visual Distortions2017
- ImageNet classification with deep convolutional neural networksCommunications of the ACM, 2017
- Deep Convolutional Neural Networks and Data Augmentation for Environmental Sound ClassificationIEEE Signal Processing Letters, 2017
- Understanding Data Augmentation for Classification: When to Warp?2016
- Deep Residual Learning for Image Recognition2016
- Accessible synthetic aperture radar system for autonomous vehicle sensing2016
- Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification2015
- Going deeper with convolutions2015
- A Dataset and Taxonomy for Urban Sound Research2014
- Representation Learning: A Review and New PerspectivesIEEE Transactions on Pattern Analysis and Machine Intelligence, 2013