基于小波变换的数字助听器关键技术研究

发布时间：2018-03-28 04:27

本文选题：数字助听器　切入点：小波阈值去噪　出处：《南京邮电大学》2017年硕士论文

【摘要】：佩戴数字助听器能提高听力患者对不同声音环境的适应能力。语音增强和响度补偿属于数字助听器的关键技术。本文在研究传统算法,以及总结前人算法的基础上,提出了对语音增强和响度补偿的改进算法,主要做了如下几方面工作:1、分析介绍了数字助听器的核心思想以及语音信号理论知识,包括语音信号的生成原理和其特征,语音信号的听觉原理;简单阐述了语音信号的三个数学模型以及主要的几种分析方法;研究了数字助听器的核心技术算法,如:声源定位、语音增强、宽动态压缩以及回馈消除算法等。重点研究了数字助听器中的语音增强和响度补偿算法。2、研究了广义旁瓣抵消器自适应波束形成算法带来的语音泄漏以及残留噪声问题。在分析语音泄漏原因的基础上,引进了小波阈值去噪技术,提出了基于小波去噪的自适应波束形成算法。小波阈值去噪能加强输入语音的信噪比,因此通过GSC结构后的输出语音更加准确。仿真结果表明,提出的算法在一定程度上减少了语音的泄漏和残留噪声,提升了语音的辨识率。3、深入研究了小波变换的响度补偿算法。现有的小波域响度补偿算法由于高频部分携带信息少,进行高增益会导致听力患者受到二次伤害,因此小波变换只对低频部分进行响度补偿。本文提出将小波域与频移技术相结合,从而实现响度补偿的算法,即把高频信号频移到低频部分,再进行响度补偿。研究表明,改进的算法可以保护语音信号的完整度,进而提高语音的舒适度和辨识率。
[Abstract]:Wearing digital hearing aids can improve the adaptability of hearing patients to different sound environments. Speech enhancement and loudness compensation are the key technologies of digital hearing aids. This paper proposes an improved algorithm for speech enhancement and loudness compensation, mainly does the following work: 1, analyzes and introduces the core idea of digital hearing aid and the theoretical knowledge of speech signal, including the principle and characteristics of speech signal generation. The hearing principle of speech signal, three mathematical models of speech signal and several main analysis methods are briefly described, and the core technology algorithms of digital hearing aid, such as sound source location, speech enhancement, are studied. The speech enhancement and loudness compensation algorithm of digital hearing aid. The speech leakage and residual noise caused by adaptive beamforming algorithm of generalized sidelobe canceller are studied. On the basis of analyzing the cause of voice leakage, The wavelet threshold denoising technique is introduced, and an adaptive beamforming algorithm based on wavelet de-noising is proposed. Wavelet threshold denoising can enhance the SNR of the input speech, so the output speech with GSC structure is more accurate. To a certain extent, the proposed algorithm reduces the leakage and residual noise of speech, improves the recognition rate of speech. 3, and deeply studies the loudness compensation algorithm of wavelet transform. High gain will result in secondary injury to hearing patients, so wavelet transform only compensates the low frequency part of loudness. In this paper, we propose a new method to realize loudness compensation by combining wavelet domain with frequency shift technique. The research shows that the improved algorithm can protect the integrity of the speech signal and improve the comfort and recognition rate of the speech.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R764.5;TN912.3

【参考文献】