Deep Convolution Network for Direction of Arrival Estimation With Sparse Prior

Wu, Liuli; Liu, Zhang‐Meng; Huang, Zhitao

doi:10.1109/lsp.2019.2945115

Cited by 148 publications

(110 citation statements)

References 15 publications

Supporting

Mentioning

110

Contrasting

Order By: Relevance

“…To balance this trade-off, we used noisy data-sets with several SNR TRAIN levels during training so that the network attains reasonable tolerance to corrupted/imperfect inputs. While similar performance degradation is also observed in [6], [7], no justification is provided for this issue.…”

Section: Numerical Simulationsmentioning

confidence: 84%

“…DOA estimation via DL is considered in [5] where a multilayer perceptron (MLP) architecture is proposed to resolve two target signals. In [6], the authors studies the same problem, also for two signal case, by exploiting the sparsity of the received signal in angular domain and design a deep convolutional neural network (CNN). In [10], a single sound source is assumed and an MLP architecture is used to estimate the source DOA angle for wideband case.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

DeepMUSIC: Multiple Signal Classification via Deep Learning

Elbir

2020

IEEE Sens. Lett.

134

View full text Add to dashboard Cite

This letter introduces a deep learning (DL) framework for direction-of-arrival (DOA) estimation. Previous works in DL context mostly consider a single or two target scenario which is a strong limitation in practice. Hence, in this work, we propose a DL framework for multiple signal classification (DeepMUSIC). We design multiple deep convolutional neural networks (CNNs), each of which is dedicated to a subregion of the angular spectrum. In particular, each CNN is fed with the array covariance matrix and it learns the MUSIC spectra of the corresponding angular subregion. We have shown, through simulations, that the proposed DeepMUSIC framework has superior estimation accuracy and exhibits less computational complexity in comparison with both DL and non-DL based techniques.

show abstract

Section: Numerical Simulationsmentioning

confidence: 84%

Section: Introductionmentioning

confidence: 99%

DeepMUSIC: Multiple Signal Classification via Deep Learning

Elbir

2020

IEEE Sens. Lett.

134

View full text Add to dashboard Cite

show abstract

“…This experiment investigates the stability of the proposed estimation network. We compare the proposed method with other methods, including MUSIC algorithm 6 , RBF-NN 28 , DNN group 37 , and DCN estimator 39 . The results are obtained based on 200 times independent Monte Carlo experiments.…”

Section: Resultsmentioning

confidence: 99%

“…This operation ignores the structure of covariance matrix and can reduce the quantity of information. In order to make full use of the information hidden in the covariance matrix, researchers also propose some deep convolution network (DCN) based methods, such as the DCN estimator used for estimating multi-DOA with sparse prior 39 , the DCN used for estimating DOAs of multi-speaker with noise signals 40 , and the de-multipath DCN models used for achieving multipath DOA estimation 41 . Although these DCN based methods use the convolutional layers to extract the deep features of covariance matrix, the pre-processing operation proposed in these articles 39 – 41 can complicate the estimation problem and cause time-consuming.…”

Section: Introductionmentioning

confidence: 99%

Multi-DOA estimation based on the KR image tensor and improved estimation network

Yuan

Yang

et al. 2021

Sci Rep

View full text Add to dashboard Cite

Deep neural networks have shown great performance for direction-of-arrival (DOA) estimation problem, but it is necessary to design some suitable networks to solve the multi-DOA estimation problem. In this paper, we use Khatri–Rao product to increase the degree of freedom of antenna array and obtain the image tensor of covariance matrix, then we propose an improved estimation network to process the tensor. We use the curriculum learning scheme and partial label strategy to develop a CurriculumNet training scheme. The training/validation results shows that the proposed training scheme can increase the generalization of the estimation network and improve the accuracy of network around $$10\%$$ 10 % . The estimation performance of the proposed network shows high-resolution results, which can distinguish two adjacent signals with angle difference of $$3^\circ $$ 3 ∘ . Moreover, the proposed estimation network has root mean square estimation error lower than $$1^\circ $$ 1 ∘ when signal noise ratio equals $$-\,10\,{\mathrm {dB}}$$ - 10 dB and can estimate DOAs precisely by only 8 snapshots, which performs much better than prior deep neural network based estimation methods and can estimate multi-DOA results under hostile estimation environments.

show abstract

“…This operation ignores the structure of covariance matrix and can reduce the quantity of information. In order to make full use of the information hidden in the covariance matrix, researchers also propose some convolution neural network (CNN) based methods, such as the deep CNN (DCNN) used for estimating multi-DOA with sparse prior 35 , the DCNN used for estimating DOAs of multi-speaker with noise signals 36 , and the de-multipath CNN models used for achieving multipath DOA estimation 37 . Although these CNN based methods use the convolutional layers to extract the deep features of covariance matrix, the pre-processing operation proposed in these articles [35][36][37] can complicate the estimation problem and cause time-consuming.…”

Section: Introductionmentioning

confidence: 99%

Multi-DOA Estimation Based on the KR Image Tensor and Improved CNN

Yuan

Yang

et al. 2021

Preprint

View full text Add to dashboard Cite

In this paper, we propose an improved convolutional neural network (CNN) to solve the multi-DOA estimation problem. We use Khatri-Rao (KR) product to obtain the KR image tensor of covariance matrix and use the proposed estimation CNN to process the tensor. In order to increase the generalization of the proposed CNN and adapt the multi-label classification problem, we use the curriculum learning scheme (CLS) and partial label strategy (PLS) to develop an efficient training procedure. We implement several experiments to demonstrate the satisfying performance of the proposed estimation method. The simulation results show that our proposed method can finish the high resolution multi-DOA estimation use only a few sensors. Furthermore, the proposed method can obtain high estimation accuracy under low SNR situations and use fewer snapshots.

show abstract

Deep Convolution Network for Direction of Arrival Estimation With Sparse Prior

Cited by 148 publications

References 15 publications

DeepMUSIC: Multiple Signal Classification via Deep Learning

DeepMUSIC: Multiple Signal Classification via Deep Learning

Multi-DOA estimation based on the KR image tensor and improved estimation network

Multi-DOA Estimation Based on the KR Image Tensor and Improved CNN

Contact Info

Product

Resources

About