MFFNet: Multi-Path Features Fusion Network for Source Enumeration

Fan, Rong; Zhu, Xinyu; Tang, Wen; Si, Chengke

doi:10.1109/lcomm.2021.3139676

Cited by 10 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This method designs a 15-layer deep neural network, which takes the covariance matrix of the spatially smoothed signal and its eigenvalues as input for training and testing. Experiment shows that this method has good estimation performance at low SNR Fan et al [24] proposed a multipath features fusion network, which fused the spatial features of the array and the temporal features of the snapshots using the multi-scale scheme of the Feature Pyramid Network and the Path Augmentation Scheme of the Path Aggregation Network. Therefore, sufficient source information can be extracted and better estimation performance can be obtained in the real environment Hu et al [25] proposed an independent source number estimation method based on the supervised learning convolutional neural networks (CNN).…”

Section: Researchers Have Proposed To Improve the Music Algorithmmentioning

confidence: 99%

An Effective Source Number Enumeration Approach Based on SEMD

Rum

Ibrahim

et al. 2022

IEEE Access

View full text Add to dashboard Cite

In signal processing, empirical mode decomposition (EMD) first decomposes the received single-channel signal into several intrinsic mode functions (IMFs) and a residual, and then uses machine learning methods for source number enumeration. EMD, however, has an end effect that can undermine the accuracy of source number enumeration. To address this issue, this paper proposed a new EMD method named Supplementary Empirical Mode Decomposition (SEMD), which improved the accuracy by extending the signal length. The proposed method can be better applied to the modal parameter identification of nonstationary and nonlinear data in the engineering field. This method first identifies two candidate extreme points, which are the closest to the function value of the first extreme point near the endpoint. Then, on one side of the candidate point, it finds a waveform similar to that at the endpoint. Finally, the maximum and minimum points at each end of the signal will be added to extend the length of the signal. The added extreme points are candidate extreme points in similar waveforms. For the improved source number enumeration method based on SEMD, the instantaneous phase is obtained first by SEMD and Hilbert transform (HT). Then, the instantaneous phase feature is extracted to obtain a high-dimensional eigenvalue vector. Finally, the back propagation (BP) neural network is used to predict the number of sources. Experiment shows that SEMD can effectively restrain the end effect, and the source number enumeration algorithm based on SEMD has a higher correct detection probability than others. 17 18 if the number of signal sources is wrong. Therefore, the 48 estimation of the number of signal sources is the primary task 49 of DOA estimation. 50 Early methods of estimating the number of signal sources 51 were based on the method of hypothesis testing, but these 52 methods may be affected by subjective behavior during man-53 ual settings. In order to avoid this problem, later researchers 54 proposed estimation methods based on information theoretic 55 criteria (ITC). Akaike information criterion (AIC) proposed 56 by Wax and Kailath [11] is a representative of these algo-57 rithms. The AIC method has good estimation performance 58 under a low signal-to-noise ratio (SNR), but the method is 59 not consistent in estimation and performs poorly at high SNR 60 and a large number of snapshots. Minimum description length 61 (MDL) proposed by Wax and Ziskind [12] achieved consis-62 tent estimation, but it performs pooly at low SNR and a small 63 number of snapshots. Later, researchers improved the source 64 number estimation method based on ITC Guo et al. [13] pro-65 posed a new source number estimation method based on 66 MDL, which uses the modified covariance matrix and its 67 eigenvectors to replace the eigenvalues used to estimate the 68 source number. A new decision variable with better anti-noise 69 ability is obtained by a series of transformations on the snap-70 shot vector and the feature vector. Experiment shows that 71 this metho...

show abstract

Section: Researchers Have Proposed To Improve the Music Algorithmmentioning

confidence: 99%

An Effective Source Number Enumeration Approach Based on SEMD

Rum

Ibrahim

et al. 2022

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The convolutional layer (Conv), full connected layer (FC), batch normalization (BN), Sigmoid, the rectified linear unit (ReLU), LeakyReLU, upsampling (Ups), global pooling, and max pooling (Pool) are referred in [18]. SR denotes the residual module with the SE operation [16]. CSR is constructed by two SR modules with CSP connections [13].…”

Section: B Yolo-doa Designmentioning

confidence: 99%

YOLO-DoA: A lightweight, real-time, and effective implementation for DoA estimation

Si¹,

Fan²

2022

Preprint

Self Cite

View full text Add to dashboard Cite

<div>In this correspondence, an efficient YOLOv3-based approach for DoA estimation named YOLO-DoA is proposed, which is implemented as a detection task to spatially separated angular boxes. DoAs of sources with confidence scores are directly predicted from the spectrum proxy with YOLO-DoA and an end-to-end estimation is realized. By combining squeeze and-excitation operation, cross stage partial connections, and an improved loss for angular box regression, the performance of YOLO-DoA is significantly enhanced.</div>

show abstract

“…A study employs Fuzzy C-Means (FCM) clustering, using Gerschgorin circle radius and its difference as features to enhance estimation [21]. Fan et al devise a network for feature fusion in multipath scenarios, aiming to enhance accuracy [22]. However, these methods require extensive training with abundant signal samples and have limited generalization capabilities.…”

Section: Introductionmentioning

confidence: 99%

An Enhanced Method Based on LSTM and Regularized K-SVD for Source Number Estimation

Pan,

Jiang,

Tian

et al. 2023

Preprint

View full text Add to dashboard Cite

In complex environments with limited snapshots, most source number estimation algorithms encounter significant challenges. To address this issue, an enhancement method is proposed involving the integration of an optimized LSTM model with the regularized K-SVD algorithm.Initially, the optimization model is proposed based on the predictability of the LSTM sequence network. This model utilizes limited data to predict additional snapshots, thereby enhancing the covariance matrix of the signal. Subsequently, a sparse reconstruction algorithm based on the regularized K-SVD method is applied to effectively filter out noise, improving signal quality. The combination of the enhancement method with source number estimation techniques further boosts the performance of traditional algorithms. Experimental analysis confirms the superiority of the proposed enhancement method in colored noise environments. Simultaneously, in experiments with a limited number of snapshots, the proposed method significantly enhances estimation performance.

show abstract

MFFNet: Multi-Path Features Fusion Network for Source Enumeration

Cited by 10 publications

References 18 publications

An Effective Source Number Enumeration Approach Based on SEMD

An Effective Source Number Enumeration Approach Based on SEMD

YOLO-DoA: A lightweight, real-time, and effective implementation for DoA estimation

An Enhanced Method Based on LSTM and Regularized K-SVD for Source Number Estimation

Contact Info

Product

Resources

About