摘要声源定位技术是一项包含多个领域的高技术课题,不仅包括声音信号的采集、AD信号转换、还包括数字信号的预处理、电子学等。从这涉及的范畴可知，这个课题的研究涉及到了许多或复杂或简单的理论知识，需要综合许多方面的知识和技术才能取得好的研究成果。近些年来，随着诸多军事和民用方面的需要及声音处理技术和信息感知技术的进展，导致声源定位技术发展的越来越迅速，应用范畴越来越广。86610

本系统采用的是基于麦克风阵列声源定位的方法，使用2个到4个麦克风接收信号，通过系统处理，可以使摄像头指向声音的来源地。这个系统是小组完成，一人负责后端设计与开发，一人负责前端设计与展示，本人负责开发实现后端部分。

后端系统部分主要负责声音的采集与预处理模块、时延估计模块和定位算法这3个模块。声音的接收部分使用麦克风序列采集声音信号，然后通过USB数据采集卡对其进行模数转换；预处理部分主要用两种方法对数字信号进行处理，分别是低通滤波和加窗分帧。时延估计部分采用广义互相关函数算法，在这模块中同时又用到了CUDA并行化的快速傅里叶变化及反变换，用来对时域和频域的相互转换，这大大优化了系统性能。定位算法部分用到了电压差技术和双曲线算法技术。

毕业论文关键词：声源定位；麦克风阵列；TDOA定位算法；CUDA优化；jestson-tk1

Abstract Sound source localization technology is a subject of involved in sound acquisition, signal detection, signal conversion ad, electronics, and other fields, thus it may be known, this topic research relates to the many theories or complex or simple, need comprehensive many aspects of knowledge and technology in order to achieve good results。 In recent years, along with the needs of many military and civil aspects and the development of sound processing technology and information sensing technology, sound source localization technology has become more and more rapid, more and more widely used in the field of application。

This system uses the microphone array sound source localization method, uses 2 to 4 microphones to receive the signal, through the system processing, may cause the camera to point to the sound source。 System is pided into the back-end design and front-end design, I am responsible for the back-end design。

The back-end system is mainly responsible for the 3 modules, which are the acquisition and preprocessing module, the time delay estimation module and the positioning algorithm module。 The acquisition part of the voice uses a microphone to receive signals, and the data acquisition card is used for digital analog conversion; the preprocessing part mainly uses the low pass filtering and the USB data acquisition card to process the digital signal。 Time delay estimation part of the generalized cross correlation function algorithm。 In this module and use the CUDA parallel fast Fourier transform and inverse transform, used in the time domain and frequency domain conversion, which can greatly improve the system performance。 In the part of the positioning algorithm, the hyperbolic algorithm and the plane square algorithm are used。

Keywords: Sound source localization; Microphone sequence; TDOA localization algorithm; CUDA optimization; Jetson-tk1

目  录

第一章 绪论 1

1。1 课题背景 1

1。2 课题目的及意义 1

1。3 课题技术现状 2

1。4 本文主要工作及论文结构 3

1。5 小结 3

第二章 声源定位系统的需求分析与功能设计