基于声传感网的多声源定位方法研究

发布时间：2018-08-05 11:34

【摘要】：本文依托国家自然科学基金(61401203)和江苏省自然科学基金(BK20130776),开展了基于声传感网的多声源定位方法研究。声传感网是综合了传感器技术、嵌入式计算和分布式信息处理等技术的一种全新信息获取和处理技术,它是由大量的具有接收声信号和计算能力的微型传感器节点构成的网络系统。声源定位是声传感网的重要应用之一,目前已在包括狙击手定位系统、野外鸟类监测系统以及视频会议等诸多军事、生态和商业领域得到了应用,因此基于声传感网的声源定位方法研究具有较高的实际应用价值,同时也是近年来信号处理领域的研究热点之一。本课题针对基于声传感网的多声源定位方法,介绍了声传感网络的概念后归纳总结并比较了三种基于声传感网的多声源定位策略,重点针对基于波达方向的交叉定位策略开展研究。首先介绍了传统的节点级声源测向方法,针对宽间距阵列形式节点在混响环境下的多声源波达方向(DOA)估计问题,基于扩散噪声模型研究了一种能够有效抑制空间混叠的迭代信噪比加权聚合角度谱方法。该方法利用声源信号的时频稀疏和短时正交性,在每个时频点针对空间混叠周期的频变特点迭代消除各真实声源在时频局域角度谱中的伪峰,从而在最终的聚合角度谱中有效抑制了空间混叠的影响。针对由多个麦克风阵列节点组成的声传感网多声源DOA交叉定位时存在的配对模糊问题,研究了一种无模糊DOA关联算法。,采用最小方差无失真响应(MVDR)波束形成方法,在每个阵列节点处重构出含有相应声源能量分布特征的时频谱,再利用尺度不变特征转换(SIFT)对不同节点中得到的频谱进行配对,从而完成多声源DOA关联。通过数据仿真和外场试验进行了性能评估,实验结果验证了所提出方法的有效性和可行性。
[Abstract]:Based on the National Natural Science Foundation of China (61401203) and the Natural Science Foundation of Jiangsu Province (BK20130776), the research of multi-source localization method based on acoustic sensor network is carried out in this paper. Acoustic sensor network is a new information acquisition and processing technology which integrates sensor technology, embedded computing and distributed information processing technology. It is a network system composed of a large number of micro sensor nodes with the ability to receive acoustic signals and compute. Sound source location is one of the important applications of acoustic sensor network. It has been used in many military, ecological and commercial fields, including sniper location system, field bird monitoring system and video conference. Therefore, the research of acoustic source location based on acoustic sensor network has high practical application value, and it is also one of the research hotspots in the field of signal processing in recent years. Aiming at the multi-source localization method based on acoustic sensor network, this paper introduces the concept of acoustic sensor network and summarizes and compares three kinds of multi-sound source localization strategy based on acoustic sensor network. This paper focuses on the research of cross location strategy based on DOA. Firstly, the traditional node-level sound source direction finding method is introduced, aiming at the problem of (DOA) estimation of the direction of arrival (DOA) of multi-source nodes in reverberation environment with wide spacing array. Based on the diffusion noise model, an iterative signal-to-noise ratio weighted aggregation method for angle spectrum is proposed, which can effectively suppress spatial aliasing. Based on the time-frequency sparsity and short-time orthogonality of sound source signals, this method iteratively eliminates the pseudo-peaks of each real sound source in the time-frequency local angle spectrum at each time frequency point in view of the frequency variation characteristics of spatial aliasing period. Thus, the effect of spatial aliasing is effectively suppressed in the final polymerization angle spectrum. In order to solve the problem of pairing ambiguity in multi-source DOA cross-location of acoustic sensor network composed of multiple microphone array nodes, a non-fuzzy DOA correlation algorithm is studied. The minimum variance distortion-free response (MVDR) beamforming method is adopted. The time-frequency spectrum with the corresponding energy distribution characteristics of sound sources is reconstructed at each array node, and the spectrum obtained from different nodes is matched by scaling invariant feature conversion (SIFT) to complete the multi-source DOA correlation. The effectiveness and feasibility of the proposed method are verified by data simulation and field test.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9;TN929.5

【相似文献】