数字助听器中自适应回波抵消算法的研究

发布时间：2018-01-07 14:03

本文关键词：数字助听器中自适应回波抵消算法的研究　出处：《哈尔滨工业大学》2015年硕士论文　论文类型：学位论文

【摘要】：助听器能帮助听力障碍患者改善听力,经过数百年的发展,助听器已经进入到了全数字时代。声音的高质量、助听器的微型化是我们所追求的目标。微型化意味着麦克风与扬声器的位置很接近,导致声源经过处理输出,在被人耳接收的同时,还有一部分会被麦克风重新拾取,形成回波。回波被不断地放大,容易形成啸叫,造成系统的不稳定,也限制了助听器的最大稳定增益,降低了患者的语言可懂度以及听力舒适度。为了解决助听器中回波造成的干扰问题,本论文将针对自适应回波抵消算法展开研究。本论文深入了解了传统的回波抵消算法LMS、RLS以及NLMS算法,这些算法的系数更新方式是逐点更新,运算量较大。考虑到数字助听器对实时性的高要求,引入了分块算法BLMS,形成NBLMS算法改变滤波器系数更新方式,降低算法复杂度。虽然NBLMS算法解决了由于逐点更新造成的系统资源浪费的问题,但是NBLMS算法存在着收敛效果不好、残留大、音质不好等问题。在此基础上,将当前时刻之前的估计误差引入归一化步长因子中,提出了NBLMS_S算法;之后在此基础上设计了两个改进算法。一是利用RLS算法中“遗忘因子”的思想对滤波器系数的更新过程做了改进,生成了NBLMS_MKS算法。另外一个算法λVS_BLMS也是基于NBLMS_S做的改进,改变了步长变化方式,在估计误差与变步长之间搭建一种类似Sigmoid表达式的非线性关系。在实现改进算法NBLMS_MKS和λVS_BLMS的设计后,将两者在PC机上进行仿真,并与最初的块处理算法NBLMS_S做性能对比。通过性能对比和运算量的分析,发现相较于NBLMS_S算法,改进算法在有更快的收敛速度和更高的收敛精度。并且,在噪声处理方面,λVS_BLMS有更好的抗噪性,更适合运用在微型化的助听器中,扩大了助听器的适用范围。
[Abstract]:Hearing aids can help people with hearing impairment to improve their hearing. After hundreds of years of development, hearing aids have entered the digital age. The high quality of sound. Miniaturization of hearing aids is our goal. Miniaturization means that the microphone is close to the speaker, resulting in the sound source being processed and output while being received by the human ear. Others will be picked up by the microphone to form an echo. The echo is constantly amplified, resulting in a screeching, resulting in instability of the system and limiting the maximum stable gain of the hearing aid. The speech intelligibility and hearing comfort of the patients were reduced. In order to solve the interference caused by echo in hearing aid. This paper will focus on adaptive echo cancellation algorithms. In this paper, we deeply understand the traditional echo cancellation algorithm LMS-RLS and NLMS algorithm, these algorithms update the coefficients point by point. Considering the high requirement of real time of digital hearing aid, the block algorithm BLMS is introduced to form the NBLMS algorithm to change the filter coefficient updating method. Reduce the complexity of the algorithm. Although the NBLMS algorithm solves the problem of system resource waste caused by point-by-point update, but the NBLMS algorithm has poor convergence effect and large residual. On the basis of this, the estimation error before the current moment is introduced into the normalized step size factor, and the NBLMS_S algorithm is proposed. Then two improved algorithms are designed. One is to improve the updating process of filter coefficients by using the idea of "forgetting factor" in RLS algorithm. NBLMS_MKS algorithm is generated. Another algorithm 位 VS_BLMS is also based on the improvement made by NBLMS_S, changing the way of step size change. A nonlinear relationship similar to the Sigmoid expression is built between the estimation error and the variable step size. After the design of the improved algorithms NBLMS_MKS and 位 VS_BLMS is implemented. The two algorithms are simulated on PC and compared with the original block processing algorithm (NBLMS_S). Through the performance comparison and computation analysis, it is found that the algorithm is compared with the NBLMS_S algorithm. The improved algorithm has faster convergence speed and higher convergence accuracy. Moreover, 位 VS_BLMS has better noise resistance in noise processing, and is more suitable for use in miniaturized hearing aids. The scope of application of hearing aid is expanded.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN713

【参考文献】