DOA Estimation of Two Targets with Deep Learning

Kase, Yuya; Nishimura, Toshihiko; Ohgane, Takeo; Ogawa, Yasutaka; Kitayama, Daisuke; Kishiyama, Yoshihisa

doi:10.1109/wpnc.2018.8555814

Cited by 56 publications

(40 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The size of the training data increases exponentially (does not scale). We should note that this is not a limitation of the proposed CNN only but rather a challenge that most DL-based approaches face, due to the need to perform exhaustive training [19], [21], [22]. In addition, under the particular scenario of low SNR, estimation of N −1 is not very accurate even for standard methods such as MUSIC, EPSRIT and R-MUSIC.…”

Section: B Mixed Number Of Sourcesmentioning

confidence: 97%

“…DL-based methods enjoy several advantages over optimization-based ones: a) after training the network no optimization is required and the solution is the result of simple operations (multiplications and additions); b) they do not require any specific tuning of parameters, in contrast to optimization-based techniques, whose solution strongly depends on the tuning of those parameters, and c) they demonstrate resilience to data imperfections, e.g., using fewer snapshots, performing well in low SNR. A deep neural network (DNN) with fully connected (FC) layers was employed in [19] for DoA classification of two targets using the signal covariance matrix. However, the reported results indicate poor DoA estimation results in high SNR.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Networks for Direction-of-Arrival Estimation in Low SNR

Papageorgiou

Sellathurai

Eldar

2021

IEEE Trans. Signal Process.

140

View full text Add to dashboard Cite

In this work, we consider direction-of-arrival (DoA) estimation in the presence of extreme noise using Deep Learning (DL). In particular, we introduce a Convolutional Neural Network (CNN) that predicts angular directions using the sample covariance matrix estimate. The network is trained from multichannel data of the true array manifold matrix in the low signalto-noise-ratio (SNR) regime. By adopting an on-grid approach, we model the problem as a multi-label classification task and train the CNN to predict DoAs across all SNRs. The proposed architecture demonstrates enhanced robustness in the presence of noise, and resilience to a relatively small number of snapshots. Moreover, it is able to resolve angles within the grid resolution. Experimental results demonstrate significant performance gains in the low-SNR regime compared to state-of-the-art methods and without the requirement of any parameter tuning in both cases of correlated and uncorrelated sources. Finally, we relax the assumption that the number of sources is known a priori and present a training method, where the CNN learns to infer their number and predict the DoAs with high confidence. The increased robustness of the proposed solution is highly desirable in challenging scenarios that arise in several fields, ranging from wireless array sensors to acoustic microphones or sonars.

show abstract

Section: B Mixed Number Of Sourcesmentioning

confidence: 97%

Section: Introductionmentioning

confidence: 99%

Deep Networks for Direction-of-Arrival Estimation in Low SNR

Papageorgiou

Sellathurai

Eldar

2021

IEEE Trans. Signal Process.

140

View full text Add to dashboard Cite

show abstract

“…The technique exhibits good adaptation to array imperfections and enhance enhance generalization to unseen cases. Kase et al [21] employed deep learning to estimate DoA and evaluates the performance in the for two narrowband signals incident on linear array. Deep neural network shows appreciable estimation accuracy.…”

Section: C) Related Workmentioning

confidence: 99%

“…Deep neural network shows appreciable estimation accuracy. The deep learning formulated for a particular case exhibits very high success rate in the same case [21]. Wan et al [22] proposed a deep learning based autonomous vehicle super resolution DoA estimation for safety driving.…”

Section: C) Related Workmentioning

confidence: 99%

An Intelligent Deep Learning-Based Direction-of-Arrival Estimation Scheme Using Spherical Antenna Array With Unknown Mutual Coupling

Famoriji

Ogundepo

2020

IEEE Access

View full text Add to dashboard Cite

Spherical antenna array (SAA) has become highly attractive where hemispherical scan coverage is required as it can provide uniform directivity in all the scan directions. Various direction-ofarrival (DoA) estimation methods suffer from different problems, such as low accuracies in mismatched conditions, high computational complexity and poor estimation in a harsh environment. Another critical concern is mutual coupling (MC) characteristics between the array elements. These problems affect the quality of the navigation signal in harsh environments. This paper presents a robust DoA estimation and mutual coupling compensation technique based on convolutional neural network (CNN) for Spherical Array. Spherical harmonic decomposition (SHD) is used to facilitate feature extraction in two sets, which contains different features about the elevation and azimuth of the source for DoA estimation. The features serve as input to the learning technique for separate estimation of elevation and azimuth, which consequently reduce computational complexity as against the joint estimation of DoA. Learning methods for DoA estimation with few frames and dense search grids within the spherical array configuration are presented. To solve the MC error, the DoA estimation scheme is also used to obtain accurate spectrum peak in the multipath scenario with unknown MC and sharper spectrum peak via the unique structure of the MC matrix and spatial smoothing algorithms. In all, experimental results, which is the ground truth to test any procedure, show the effectiveness, validity, and potential practical application of the proposed technique.

show abstract

“…Two formats of data are used in the process: the raw radar data cube and the statistical average in the form of covariance matrices. Simulation and experiment results are provided to demonstrate the superior performance of the proposed method over conventional methods such MUSIC as well as emerging deep learning methods [ 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 ]. It has been claimed, based on simulation results, that neural networks (NNs) can outperform MUSIC when SNRs are relatively low; however, no convincing experimental results have been shown to support the claims [ 14 , 15 , 16 , 17 , 18 , 19 , 20 ], and the angular differences between targets are not small enough to show the supremacy of deep-learning-based methods [ 20 ].…”

Section: Introductionmentioning

confidence: 99%

DoA Estimation for FMCW Radar by 3D-CNN

Sang

Chien

Chang

et al. 2021

Sensors

View full text Add to dashboard Cite

A method of direction-of-arrival (DoA) estimation for FMCW (Frequency Modulated Continuous Wave) radar is presented. In addition to MUSIC, which is the popular high-resolution DoA estimation algorithm, deep learning has recently emerged as a very promising alternative. It is proposed in this paper to use a 3D convolutional neural network (CNN) for DoA estimation. The 3D-CNN extracts from the radar data cube spectrum features of the region of interest (RoI) centered on the potential positions of the targets, thereby capturing the spectrum phase shift information, which corresponds to DoA, along the antenna axis. Finally, the results of simulations and experiments are provided to demonstrate the superior performance, as well as the limitations, of the proposed 3D-CNN.

show abstract

DOA Estimation of Two Targets with Deep Learning

Cited by 56 publications

References 6 publications

Deep Networks for Direction-of-Arrival Estimation in Low SNR

Deep Networks for Direction-of-Arrival Estimation in Low SNR

An Intelligent Deep Learning-Based Direction-of-Arrival Estimation Scheme Using Spherical Antenna Array With Unknown Mutual Coupling

DoA Estimation for FMCW Radar by 3D-CNN

Contact Info

Product

Resources

About