水声信号检测与识别中信号处理和特征约简的新方法研究

It is general method of the use of underwater acoustic signal for the detection of underwater target in most of researches. But with the development of anti-noise technology and electronic warfare technology the existing detection technology of underwater acoustic signal can’t meet the objective demand in modern intelligent ocean target detection and recognition, so it is urgent to develop an efficient and intelligent detection and recognition technology of underwater acoustic signal. In this project, from the perspective of reduction and synergic analysis of underwater acoustic signal we suggest to utilize signal processing, sparse representation and convex optimization algorithm to investigate denoising/interference excision, fusion of detection results, dimension reduction/feature selection, and selective combination of heterogeneous feature sets in the detection and recognition of underwater acoustic signal. Finally we aim to construct an efficient and reliable detection and recognition model of underwater acoustic signal that enable effective detection and recognition of underwater targets. In the research of the project not only it could preserve the advantages and characteristics of the detection and recognition of underwater acoustic signal, but also it could provide robust solution for the fusion of detection results and the combination of heterogeneous feature sets of underwater acoustic signal which can improve the detection and recognition performance of underwater targets. In summary, this study will provide new ideas and new methods in the field of underwater acoustic detection and recognition for our country, which have important academic value and application prospect.

利用水声信号对水下目标检测与识别是目前国内外通用的手段，但随着抗噪音技术和电子对抗技术的发展，现有的水声信号检测技术已不能满足现代海洋目标智能化检测的客观需求，迫切需要发展一种高效和智能化水声信号检测与识别技术。本项目研究拟利用信号处理、稀疏表示和凸优化算法，从水声信号约简与协同分析的角度出发，重点研究水声信号检测与识别中的去噪声/干扰、检测结果融合、特征维数约简/特征筛选和异质水声信号或特征的选择性融合模型与方法，建立高效、可靠的水声信号检测与识别模型，从而实现对水中目标有效检测与识别。本项目的研究在保持原有水声信号检测与识别优势和特点的基础上，能够为水声信号检测结果融合和异质水声信号特征的选择性融合问题提供鲁棒的解决方案，从而提高目标检测与识别能力。本研究将为我国水声检测与识别领域提供新思路和新方法，具有重要的学术研究价值和应用前景。

本项目研究中利用不等间距阵元进行湖上、海上试验，在对不同频带（频率）信号检测中进行了去噪声/干扰研究，并对检测结果进行了特征提取和数据融合研究；同时利用信号处理、稀疏表示和凸优化算法从特征子空间出发，研究了特征维数约简/特征筛选和异质水声信号或特征的选择性融合模型与方法，建立了高效、可靠的水声信号检测与识别模型，实现了对水中目标有效检测与识别。本项目的研究在保持原有水声信号检测与识别优势和特点的基础上，能够为水声信号检测结果融合和异质水声信号特征的选择性融合问题提供鲁棒的解决方案，从而提高目标检测与识别能力。本项目历时3年，预期的研究工作进展顺利，分别研究并分析时/频域噪声/干扰抑制和补偿方法，提出多干扰多目标下的全方位目标检测方法；研究基于特征子空间所具有的特征信息融合技术，建立基于特征子空间方位稳定性的极值加权目标最佳检测理论框架。.通过项目的开展，与国内、外学者开展了广泛且深入的学术交流，培养硕士、博士研究生4名，发表文章、专利、软件著作权等16份，其中2篇文章被SCI源刊录用，7篇文章被EI源刊录用，1篇文章修改后录用。.项目经费来自于国家自然科学基金委，下拨经费合计23.3万元。至项目结题时止，经费基本使用完毕，节余2.82万元。