(硕士论文)面向商务智能的采购决策支持系统应用研究.pdf

华中科技大学硕士学位论文
摘要
随着语音通信技术的发展,对声源定位的实时性和精确性要求越来越高,基于麦
克风阵列的语音处理方法已成为研究热点之一。然而,受系统灵活性、应用方便性和
算法精确性等方面的影响,目前国内麦克风阵列产品虽有部分应用,但是仍然具有一
定的局限性。因此,为了推进麦克风阵列的应用,对麦克风阵列及其算法开展研究,
具有实用价值和意义。
为了利用麦克风阵列实现对实际声源的定位功能,本论文着重于声源定位算法研
究。在 LabVIEW 环境中设置了仿真模型,考虑室内混响和噪声的影响,分别实现了
基于时延估计的声源定位法和基于子空间技术的声源定位法,并讨论了各种算法的适
用范围。基于时延估计的声源定位法分为时延估计和声源定位两部分,其中时延估计
部分讨论了各种加权因子的 GCC 时延估计法,声源定位部分讨论了角度距离定位法、
最小二乘法和球形插值法的特点和定位效果。基于子空间技术的声源定位讨论了二维
MUSIC 方法对远场声源的 DOA 估计。经仿真结果对比分析,讨论的各种算法基本上
可以实现声源定位功能,但是局限性和精确性各有不同,如正常情况下的最小二乘法
和球形插值法精确性好,但是不适用于四元阵列,需针对实际系统进行改进;角度距
离定位法在远场情况下对距离的估计误差较大等等。
通过仿真实现声源定位功能后,针对系统构建,将算法移植到硬件平台上,对实
际声源进行定位。在仿真过程对实际系统的理论指导下,选用了各种加权的 GCC 时
延估计法和角度距离声源定位法进行实际声源定位,实验证明,该系统能够比较精确
地对半空间域的声源进行定位。

关键词:麦克风阵列时延估计声源定位 MUSIC算法 LabVIEW
I
华中科技大学硕士学位论文
Abstract
With the development of the munications technology, attention is mainly
paid to the performance of the real-time character and the accuracy for acoustic source
localization. Thus the acoustic signal processing methods based on microphone array have
e a hotspot in the research area. However, the applications of microphone array
commercial products are still limited in domestic up to now owing to lacking the flexibility
of system, the convenience in use and the accuracy of algorithm. Therefore, in order to
popularize the application of the microphone array, the improvement of the algorithm and
hardwares on the performance of acoustic source localization are of important value.
The research is focused on the algorithm to obtain the sound source localization
practically. A simulation model is put forwad firstly in the LabVIEW environment, after
taking both the reverberation and noise into account. Then, the time delay estimation and
the sub-spatial algorithm of the acoustic source localization are carried out. As a result, the
application scope and the effects of various algorithm are further discussed. As the time
delay estimation algorithm is divided into two steps, kinds of wei