Diversified radar micro-Doppler simulations as training data for deep residual neural networks

Seyfioğlu, Mehmet Saygın; Erol, Barış; Gürbüz, Sevgi Zübeyde; Amin, Moeness G.

doi:10.1109/radar.2018.8378629

Cited by 29 publications

(13 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…as [19][20][21][22][23] make use of simulated radar returns via MoCapbased data. Even though the accuracy of the MoCap data cannot be guaranteed as some noise exists in positioning the joints during motion as the markers are considered to be affixed on a rigid structure where in fact they are fitted on skin or clothing that stretch during movements.…”

Section: Simulation Setup and Numerical Resultsmentioning

confidence: 99%

Human Activity Classification With Radar: Optimization and Noise Robustness With Iterative Convolutional Neural Networks Followed With Random Forests

et al. 2018

View full text Add to dashboard Cite

The accurate classification of activity patterns based on radar signatures is still an open problem and is key to detect anomalous behavior for security and health applications. This paper presents a novel iterative convolutional neural networks strategy with an autocorrelation pre-processing instead of the traditional micro-Doppler image pre-processing to classify activities or subjects accurately. The proposed strategy uses an iterative deep learning framework for the automatic definition and extraction of features. This is followed by a traditional supervised learning classifier to label the different activities. Using three human subjects and their real motion captured data, twelve thousand radar signatures were simulated by varying additive white Gaussian noise. Additionally, 6720 experimental radar signatures were captured with a frequency-modulated continuous radar at 5.8GHz with 400M Hz of instantaneous bandwidth from seven activities using one subject and 4800 signatures from five subjects while walking. The simulated and experimental data were both used to validate our proposed method. With SNR varying from −20 to 20dB with 88.74% average accuracy at −10dB and 100% peak accuracy at 15dB. The proposed Iterative Convolutional Neural Networks followed with Random Forests (ICNNRF) does not only outperform the feature-based methods using micro-Doppler images but also the classification methods using other types of supervised classifiers after our proposed iterative convolutional neural network.

show abstract

Section: Simulation Setup and Numerical Resultsmentioning

confidence: 99%

Human Activity Classification With Radar: Optimization and Noise Robustness With Iterative Convolutional Neural Networks Followed With Random Forests

et al. 2018

View full text Add to dashboard Cite

show abstract

“…Furthermore, the performance of ResNet18 that is trained with simulated radar data is the best, with an average F1 score of 97.1%. Based on these results, we have reasons to believe that using simulated radar data to train the feature extractor has more potential to achieve good performance for HMR [19], [39].…”

Section: B Ablation Study On Csdsmentioning

confidence: 94%

Human Motion Recognition With Limited Radar Micro-Doppler Signatures

Fioranelli

et al. 2021

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

The performance of deep learning (DL) algorithms for radar-based human motion recognition (HMR) is hindered by the diversity and volume of the available training data. In this article, to tackle the issue of insufficient training data for HMR, we propose an instance-based transfer learning (ITL) method with limited radar micro-Doppler (MD) signatures, alleviating the burden of collecting and annotating a large number of radar samples. ITL is a unique algorithm that consists of three interconnected parts, including DL model pretraining, correlated source data selection, and adaptive collaborative fine-tuning (FT). Any of the three components cannot be excluded; otherwise, the performance of the entire algorithm decreases. The experiments with a radar data set of six human motions show that ITL achieves state-of-the-art performance for HMR with limited training samples, outperforming several existing transfer learning approaches. Especially, when there are only 100 samples per person per class, ITL yields an F1 score of 96.7%. Last but not least, ITL is more generalized to human motion differences. Though adapted to recognize the persons' motions in a small-scale target data set, ITL can also classify the persons' motion data used for pretraining, achieving up to 11.0% F1 score enhancement over the conventional FT method.

show abstract

“…The decision about which type of transfer learning has to be preferred is based on the size of the available dataset and on the similarity of the last dataset with the one used for pre-training the selected network architecture. It has been shown that the final score of a DCNN-based classifier can be improved either by exploiting a pretraining procedure based on a simulated radar dataset [139] or by employing a pre-trained DCNN on a separate large scale RGB dataset [138].…”

Section: A Transfer Learningmentioning

confidence: 99%

Machine Learning and Deep Learning Techniques for Colocated MIMO Radars: A Tutorial Overview

DAVOLI¹,

Guerzoni²,

Vitetta³

2021

Preprint

View full text Add to dashboard Cite

<p>Radars are expected to become the main sensors in various civilian applications, ranging from health-care monitoring to autonomous driving. Their success is mainly due to the availability of both low cost integrated devices, equipped with compact antenna arrays, and computationally efficient signal processing techniques. An increasingly important role in the field of radar signal processing is played by machine learning and deep learning techniques. Their use has been first taken into consideration in human gesture and motion recognition, and in various healthcare applications. More recently, their exploitation in object detection and localization has been also investigated. The research work accomplished in these areas has raised various technical problems that need to be carefully addressed before adopting the above mentioned techniques in real world radar systems. In this manuscript, a comprehensive overview of the machine learning and deep learning techniques currently being considered for their use in radar systems is provided. Moreover, some relevant open problems and current trends in this research area are analysed. Finally, various numerical results, based on both synthetically generated and experimental datasets, and referring to two different applications are illustrated. These allow readers to assess the efficacy of specific methods and to compare them in terms of accuracy and computational effort.</p>

show abstract

Diversified radar micro-Doppler simulations as training data for deep residual neural networks

Cited by 29 publications

References 16 publications

Human Activity Classification With Radar: Optimization and Noise Robustness With Iterative Convolutional Neural Networks Followed With Random Forests

Human Activity Classification With Radar: Optimization and Noise Robustness With Iterative Convolutional Neural Networks Followed With Random Forests

Human Motion Recognition With Limited Radar Micro-Doppler Signatures

Machine Learning and Deep Learning Techniques for Colocated MIMO Radars: A Tutorial Overview

Contact Info

Product

Resources

About