School of Naval Architecture and Ocean Engineering, Jiangsu University of Science and Technology, Zhenjiang 212100, China.
Sensors (Basel). 2023 May 19;23(10):4919. doi: 10.3390/s23104919.
Acoustic vector sensor (AVS) is a kind of sensor widely used in underwater detection. Traditional methods use the covariance matrix of the received signal to estimate the direction-of-arrival (DOA), which not only loses the timing structure of the signal but also has the problem of weak anti-noise ability. Therefore, this paper proposes two DOA estimation methods for underwater AVS arrays, one based on a long short-term memory network and attention mechanism (LSTM-ATT), and the other based on Transformer. These two methods can capture the contextual information of sequence signals and extract features with important semantic information. The simulation results show that the two proposed methods perform much better than the multiple signal classification (MUSIC) method, especially in the case of low signal-to-noise ratio (SNR), the DOA estimation accuracy has been greatly improved. The accuracy of the DOA estimation method based on Transformer is comparable to that of the DOA estimation method based on LSTM-ATT, but the computational efficiency is obviously better than that of the DOA estimation method based on LSTM-ATT. Therefore, the DOA estimation method based on Transformer proposed in this paper can provide a reference for fast and effective DOA estimation under low SNR.
声矢量传感器（AVS）是一种广泛应用于水下探测的传感器。传统方法使用接收信号的协方差矩阵来估计到达方向（DOA），这不仅丢失了信号的定时结构，而且存在抗噪声能力弱的问题。因此，本文提出了两种用于水下 AVS 阵列的 DOA 估计方法，一种基于长短期记忆网络和注意力机制（LSTM-ATT），另一种基于 Transformer。这两种方法可以捕获序列信号的上下文信息，并提取具有重要语义信息的特征。仿真结果表明，所提出的两种方法比多重信号分类（MUSIC）方法性能更好，特别是在低信噪比（SNR）情况下，大大提高了 DOA 估计精度。基于 Transformer 的 DOA 估计方法的精度可与基于 LSTM-ATT 的 DOA 估计方法相媲美，但计算效率明显优于基于 LSTM-ATT 的 DOA 估计方法。因此，本文提出的基于 Transformer 的 DOA 估计方法可以为低 SNR 下快速有效的 DOA 估计提供参考。
