基于麦克风阵列的语音增强方法研究

发布时间：2018-06-15 22:41

本文选题：广义互相关 + 双四元十字阵　；参考：《东北石油大学》2017年硕士论文

【摘要】：语音增强是一项应用广泛的研究,经过几十年的发展已经深入到人们生活的很多方面,尤其伴随手机的普及,可以说语音增强研究的成果已经时刻伴随着我们每一个人,影响着每一个人的生活。所以,对于语音增强算法的研究仍然是一个重要的命题。本文题目包含了研究的两个方面,分别是麦克风阵列以及进行去噪的滤波算法。麦克风阵列是声源定位以及基于麦克风阵列的语音增强系统的基础。通过对前人阵列研究成果的研究总结,本文对几种常用的麦克风阵列进行了对比研究,主要包括平面四元十字阵、五元十字阵,并在研究的基础设计出了双四元麦克风阵列。本文的第四章主要介绍自适应滤波算法,自适应滤波算法是本文研究的重点,本文主要学习研究最小均方算法(Least mean square, LMS),变分模态分解(variational mode decomposition,VMD)算法,并在研究的基础上提出了改进的归一化最下均方算法(Normalized Least Mean Square,NLMS)算法,通过对其仿真,表明修正的归一化LMS算法收敛速度比LMS算法快,而且失真极小,该算法相对于其他自适应算法而言具有更好的实用性。最后通过实验验证了双四元十字阵比单四元十字阵具有更高的定位精度和可靠性。实验主要包含软件设计,硬件平台的搭建。实验主要利用双四元十字阵接收语音信号,语音信号通过采集卡到达LabVIEW设计上位机软件,最后经过自适应算法处理得到相应的结果,通过实验结果和理论结果对比验证了双四元麦克风阵列的可靠性,具有一定实用价值。
[Abstract]:Speech enhancement is a widely used research. After decades of development, it has gone deep into many aspects of people's lives, especially with the popularity of mobile phones. It can be said that the results of speech enhancement research have been accompanied by every one of us all the time. It affects everyone's life. Therefore, the research of speech enhancement algorithm is still an important proposition. The topic of this paper includes two aspects: microphone array and denoising filtering algorithm. Microphone array is the basis of sound source location and speech enhancement system based on microphone array. In this paper, several commonly used microphone arrays are compared and studied, including planar quaternion cross array and five-element cross array. Based on the research, a dual-quaternion microphone array is designed. The fourth chapter of this paper mainly introduces the adaptive filtering algorithm. The adaptive filtering algorithm is the focus of this paper. In this paper, we mainly study the least mean square (LMS) algorithm, the variational mode decomposition (VMD) algorithm. An improved Normalized least mean Square algorithm (Normalized least mean Square) algorithm is proposed on the basis of the research. The simulation results show that the modified normalized LMS algorithm converges faster than the LMS algorithm and the distortion is minimized. This algorithm is more practical than other adaptive algorithms. Finally, the accuracy and reliability of double quaternion cross array are proved to be higher than that of single quaternion cross array. The experiment mainly includes software design and hardware platform. The experiment mainly uses the double quaternion cross array to receive the speech signal, and the speech signal arrives at LabVIEW through the acquisition card to design the upper computer software. Finally, the corresponding result is obtained through the adaptive algorithm processing. The reliability of the dual quaternion microphone array is verified by comparing the experimental results with the theoretical results.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN912.35

【参考文献】