【作者】 张力；

【导师】 王旭；

【作者基本信息】 东北大学 ， 电子与通信工程（专业学位）， 2013， 硕士

【摘要】 全世界约有3.6亿人患有残疾性听力障碍,耳聋已成为五大残疾之首。电子耳蜗是目前唯一能有效恢复神经性耳聋的医疗装置。其中,语音编码策略是影响电子耳蜗性能的关键的技术之一,近年来一直是电子耳蜗相关技术的研究热点。研究语音编码策略有助于进一步提高电子耳蜗的性能,对于聋人的康复事业有着重要贡献。本文在阅读了大量国内外相关文献的基础上,掌握了电子耳蜗语音编码策略的基本原理,重点研究了一种基于人耳特性的电子耳蜗语音编码策略。本文的主要工作如下：首先在分析人耳生理结构和功能作用的基础上,研究了耳蜗基底膜的特性,设计了Gammatone滤波器完成基底膜的分频作用；同时又研究了内毛细胞的特点,设计了Meddis数学模型完成了内毛细胞对信号的包络检测功能,在Simulink平台下完成了上述两种数学模型的仿真,仿真结果表明模型具有较好的抗噪性。接着在上述耳蜗数学模型的基础上,研究了一种基于人耳特性的心理声学模型的编码策略。该编码策略属于N-of-M型编码策略的一种,针对编码策略的各个阶段,分别结合人耳特性：其中带通滤波和包络检测采用了耳蜗系统数学模型中Gammatone滤波器组和Meddis内毛细胞模型；在通道选择阶段引入心理声学的掩蔽效应模型,从M个通道中选择出N个通道,在相应的通道产生脉冲刺激信号。仿真结果表明了基于人耳特性的心理声学模型的编码策略更符合人耳的听觉特性,同时在抗噪性和稳定性方面要优于CIS编码策略和ACE编码策略。最后,本文以TMS320VC5402为核心搭建了DSP硬件系统平台,使用TLV320AIC23作为语音信号采集模块,采用压频转换器VFC32产生刺激脉冲,并设计了其它相关外围电路,进而在硬件系统上实现了电子耳蜗编码策略,通过实验测试,进一步验证了该策略的正确性和合理性。更多还原

【Abstract】 There are about 360 million people suffering from hearing loss disability. Deafness has become the primary disability. Cochlear Implant(CI) is an effective medical device which can cure nerve deafness. The speech coding strategy is a key technology that can affect the performance of cochlear implant. The research of speech coding strategy is helpful to improve the CI’s performance, and contribute to the rehabilitation of the deaf patients.Many related literatures have been read in this paper. After mastered the basic principle of speech coding strategy, a new speech coding strategy based on auditory characteristics is proposed in this paper. The main work is as follows:First, based on the analysis of ear’s physiological structure and function, I have studied the characteristics of basement membrane, and designed Gammatone filter to complete frequency resolution. Besides, I also studied the characteristics of inner hair cells, and designed Meddis Model to realize envelope detection. In the Simulink environment I have completed the simulation of the two kinds of mathematical model, results show that the model has a good robust.Then, on the basis of the cochlea model, a coding strategy based on auditory characteristics with psychoacoustic model is proposed. This strategy is an N-of-M type coding strategy. In this strategy Gammatone filter bank and Meddis model were used, while at bands selection stage adding in a psychoacoustic masking effects model to choose N bands from M bands, and generate stimulation pulses in the selected bands. Simulation results show that new strategy’s stability and robust performance are better than the CIS and ACE strategies.Finally, in this paper TMS320VC5402 chip is used to built the DSP hardware system, TLV320AIC23 is designed as speech signal acquisition module, and voltage to frequency converter VFC32 is used to generate stimulation pulses. The other relevant peripheral circuits are designed for the system. The speech coding strategy based on auditory characteristics with psychoacoustic model is implemented on the DSP hardware system sucessful. The experimental results show the correctness and rationality of this strategy.更多还原