张天帅,刘金涛,王良

• 1:中国海洋大学信息科学与工程学部

摘要(Abstract)：

针对传统声纳处理器算力受限，能效比低，难以支撑水声目标识别实时推理的问题，本文基于异构SoC平台设计了面向被动声纳水下目标实时计算处理系统。该系统具有较低资源开销和较小分类精度损失等优点，是一种低时延、高能效比的硬件加速器解决方案。本文以MobileNetV1网络模型为基础并对其进行结构优化，在现场可编程门阵列(Field programmable gate array, FPGA)上通过并行流水线的加速结构实现它的前向推理过程，并对其权值参数进行二值化的处理，以达到减少存储量和计算量的同时加快其推理速度的目的。同时，根据在输入通道维度以及输出图像高度上分块并行的优化思想，设计了深度可分离卷积的流水优化策略，采用并行流水的结构极大减少了前向推理的时间。实验表明，在利用出海实际采集得到的水声数据集上，本文实现的系统识别精度为88.5%,在的分辨率的图像上，时间延迟达到4.23 ms。对比CPU速度提升了70.68倍，是GPU速度的68%。能效比分别为CPU的0.08%,GPU的2.12%。本文为神经网络在硬件资源有限以及功耗存在限制的轻量型移动端或者边缘设备上的应用与部署，以及对促进融合水下勘探网络的建设和水下信息的快速获取提供了设计思路。

关键词(KeyWords)： 水声目标分类;深度可分离卷积;定点量化;FPGA

Abstract:

Keywords:

基金项目(Foundation): 国家自然科学基金项目(52001296)资助~~

作者(Author): 张天帅,刘金涛,王良

DOI: 10.16441/j.cnki.hdxb.20230110

参考文献(References)：

• [1] 华贤兵.水下目标信号特征提取及识别技术的研究[D].南京：江苏科技大学，2010.Hua X B,Research on Feature Extraction and Recognition Techniques for Underwater Target Signals[D].Nanjing:Jiangsu University of Science and Technology,2010.
• [2] 程玉胜，张宝华，高鑫，等.船舶辐射噪声解调谱相位耦合特性与应用[J].声学学报，2012,37(1):25-29.Cheng Yusheng,Zhang Baohua,Gao Xin,et al.Phase-coupling characteristics of ship radiated-noise demodulation[J].声学学报，2012,37(1):25-29.
• [3] Wang S,Zeng X.Robust underwater noise targets classification using auditory inspired time-frequency analysis[J].Applied Acoustics,2014,78:68-76.
• [4] 郝宇昕.基于深度神经网络的舰船辐射噪声特征分类技术[D].哈尔滨：哈尔滨工程大学，2019.Hao X Y.Classification Technology of Ship Radiated Noise Features Based on Deep Neural Networks[D].Harbin:Harbin Engineering University,2019.
• [5] Leal N,Leal E,Sanchez G.Marine vessel recognition by acoustic signature[J].ARPN Journal of Engineering and Applied Sciences,2015,10(20):9633-9639.
• [6] Wang W,Li S,Yang J,et al.Feature extraction of underwater target in auditory sensation area based on MFCC[C]//IEEE:2016 IEEE/OES China Ocean Acoustics (COA).2016:1-6.
• [7] Zhang S K,Tian D.Intelligent classification method of Mel frequency cepstrum coefficient for underwater acoustic targets[J].Journal of Applied Acoustics,2019,38(2):267-272.
• [8] Zhang W,Yang X,Leng C,et al.Modulation recognition of underwater acoustic signals using deep hybrid neural networks[J].IEEE Transactions on Wireless Communications,2022,21(8):5977-5988.
• [9] Priyadharsini R,Sree S T,Rajendran V.A wavelet transform based contrast enhancement method for underwater acoustic images[J].Multidimensional Systems and Signal Processing,2018,29:1845-1859.
• [10] Kamal S,Mujeeb A,Supriya M H.Novel class detection of underwater targets using self-organizing neural networks[C]//IEEE:2015 IEEE Underwater Technology (UT).2015:1-5.
• [11] Dong Y,Shen X,Jiang Z,Wang H.Recognition of imbalanced underwater acoustic datasets with exponentially weighted cross-entropy loss[J].Applied Acoust,2021,174:107740.
• [12] Li J H,Yang H H.Deep convolutional neural networks for underwater acoustic target time-frequency feature extraction and recognition[J].Unmanned Systems Technology,2019,2(4) :26-30.
• [13] Mohammad K,Mosavi.Classification of underwater acoustical dataset using neural network trained by Chimp Optimization Algorithm[J].Applied Acoust,2020,157:107005.
• [14] Courbariaux M,Bengio Y,David J P.Binaryconnect:Training deep neural networks with binary weights during propagations[J].Advances in Neural Information Processing Systems,2015,28.
• [15] Liang S,Yin S,Liu L,et al.FP-BNN:Binarized neural network on FPGA[J].Neurocomputing,2018,275:1072-1086.
• [16] Rastegari M,Ordonez V,Redmon J,et al.Xnor-net:Imagenet classification using binary convolutional neural networks[C]//Computer Vision-ECCV 2016:14th European Conference,Amsterdam,The Netherlands,October 11-14,2016,Proceedings,Part Ⅳ.Cham:Springer International Publishing,2016:525-542.
• [17] Howard A G,Zhu M,Chen B,et al.Mobilenets:Efficient convolutional neural networks for mobile vision applications[J].DOI:arXiv preprint arXiv:1704.04861,2017.
• [18] Chollet F.Xception:Deep learning with depthwise separable convolutions[C][s.l.]//Proceedings of the IEEE conference on computer vision and pattern recognition,2017:1251-1258.
• [19] Wu D,Zhang Y,Jia X,et al.A high-performance CNN processor based on FPGA for MobileNets[C]//IEEE:2019 29th International Conference on Field Programmable Logic and Applications (FPL),2019:136-143.
• [20] Nan T,Zhu Y,Li W,et al.An FPGA-based Hardware Acceleration for Key Steps of Facet Imaging Algorithm[C]//IEEE:2019 IEEE International Conference on Smart Cloud (SmartCloud),2019:86-91.
• [21] Lin W,Yuan B,Wu P.The design of the control system of navigation for UUVin underwater information warfare[M].Xi’an:Advances in Intelligent Systems and Computing China,2013:219-226.
• [22] Doan V S,Huynh T T,Kim D S.Underwater acoustic target classification based on dense convolutional neural network[J].IEEE Geoscience and Remote Sensing Letters,2020,19:1-5.
• [23] Sifre L,Mallat S.Rigid-motion scattering for texture classification[J].Computer Science,DOI:arXiv preprint arXiv:1403.1687,2014.
• [24] Montavon G,Samek W,Müller K R.Methods for interpreting and understanding deep neural networks[J].Digital Signal Processing,2018,73:1-15.
• [25] Courbariaux M,Hubara I,Soudry D,et al.Binarized neural networks:Training deep neural networks with weights and activations constrained to +1 or -1[J].Computer Science,DOI:arXiv preprint arXiv:1602.02830,2016.
• [26] Simons T,Lee D J.A review of binarized neural networks[J].Electronics,2019,8(6):661.
• [27] Luo X,Feng Y,Zhang M.An underwater acoustic target recognition method based on combined feature with automatic coding and reconstruction[J].IEEE Access,2021(9):63841-63854.
• [28] Loshchilov I,Hutter F.Sgdr:Stochastic gradient descent with warm restarts[J].Doi:https://doi.org//10.48550/arxiv.1608.03983.
• [29] Zhang M,Li L,Wang H,et al.Optimized compression for implementing convolutional neural networks on FPGA[J].Electronics,2019,8(3):295.
• [30] Gschwend D.Zynqnet:An fpga-accelerated embedded convolutional neural network[J].Computer Science,Doi:https://doi.org//10.48550/arxiv.2005 .06892.
• [31] Kim J H,Lee J,Anderson J H.FPGA architecture enhancements for efficient BNN implementation[C]//IEEE:2018 International Conference on Field-Programmable Technology (FPT),2018:214-221.
• [32] 刘峰，罗再磊，沈同圣，等.时频谱图和数据增强的水声信号深度学习目标识别方法[J].应用声学，2021,40(4):518-524.Liu F,Luo Z L,Shen T S,et al.Deep learning target recognition method for underwater acoustic signals based on time-frequency spectrum and data augmentation[J].Applied Acoustics,2021,40(4):518-524.

文章评论(Comment)：