Efficient 2D-DOA Estimation Based on Triple Attention Mechanism for L-Shaped Array

Sensors, Journal Year: 2025, Volume and Issue: 25(8), P. 2359 - 2359

Published: April 8, 2025

Accurate direction-of-arrival (DOA) estimation is crucial to a variety of applications, including wireless communications, radar systems, and sensor arrays. In this work, we propose novel deep convolutional neural network (DCN) called TADCN for 2D-DOA using an L-shaped array. The achieves high performance through triple attention mechanism (TAM). Specifically, the new architecture enables capture relationships across channel, height, width dimensions signal sample features, thereby enhancing feature extraction capability improving resulting spatial spectrum. To end, spectrum processed by proposed analyzer yield high-precision DOA results. An automatic angle matching method based on employed estimating pairing between estimated azimuth elevation sets. Furthermore, overall efficiency enhanced parallel processing networks. Simulation results demonstrate that algorithm outperforms traditional methods learning-based approaches various noise levels snapshots while maintaining better even in presence correlated sources.

Language: Английский

---

Twin proximal support vector regression with heteroscedastic Gaussian noise

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 250, P. 123840 - 123840

Published: March 26, 2024

Language: Английский

---

Direction of Arrival Estimation Based on DNN and CNN

Electronics, Journal Year: 2024, Volume and Issue: 13(19), P. 3866 - 3866

Published: Sept. 29, 2024

The accuracy of Direction Arrival (DOA) estimation primarily depends on the precision data. When receiver uses a low-precision analog-to-digital converter (ADC), traditional DOA algorithms exhibit poor accuracy. To face challenge multi-target in scenarios with ADC quantized sampling, this paper proposes novel algorithm for signals based classification problems. A deep learning network was constructed using Deep Neural Networks (DNNs) and Convolutional (CNNs), divided into signal recovery framework framework. DNN is utilized to recover that have undergone quantization, while CNN addresses problem estimate from received data an unknown number sources. comprehensive analysis impact signal-to-noise ratio (SNR), array elements, quantization bits proposed conducted. Simulation results indicate exhibits superior performance scenarios, characterized by reduced computational complexity, thereby facilitating real-time estimation.

Language: Английский

---

Soft sensor modeling of steel pickling concentration based on IGEP algorithm

Canadian Metallurgical Quarterly, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 13

Published: Nov. 26, 2024

Accurate measurement of acid concentration is paramount for ensuring the quality strip steel pickling. Online measurement, a method that reduces operational complexity and lags effectively, gradually replacing offline concentration. In this study, an indirect soft sensor model based on improved gene expression programming (IGEP) algorithm has been constructed, leveraging easily measurable indexes from large-scale dataset. The IGEP-based predicted mean absolute errors H+ Fe2+ concentrations were 1.72 1.98 g/L, respectively. Additionally, goodness fit values prediction models 0.945 0.933, Compared with support vector regression (SVR), which suitable small samples, it was demonstrated achieved better predictive performance. Taken together, our study designed more effective practical determining pickling, providing new ideal choice industry, profound significance in control pickling solution production steel.

Language: Английский

---