Study on a compound hearing aid progressing based on speech analysis and synthesis

我们开发了一个基于语音分析和合成模型的复合式助听器系统。该系统是一种新的助听器，可以选择性地或同时执行各种常规的助听器处理。在1999年，我们研究了一个完整的算法理论。特别是，证实了频率扭曲和电平压缩对基音转换语音的自然度有影响。结果，它已经能够定义频率翘曲系数和音调转换比之间的函数。2000年，我们研究了一种语音速率控制算法，它由局部转换和全局时间转换两部分组成。而VUS决定，这是重新合成的援助演讲质量的一个重要因素，我也进行了调查。并在WINDOWS平台上实现了评价系统。因此，它已经提供了一套在线评估测试。然而，目前还没有在DSP系统上实现实时仿真。未来，我们计划在已经购买的一块DSP芯片的基础上，设计一个基于I a板的实时操作系统。

We have developed a compound hearing aid system based on a speech analysis and synthesis model. The system is a new hearing aid one to be executed various conventional hearing aid processing selectively or simultaneously. In 1999, we have investigated a whole algorithm theoreticauy. Especially, it was confirmed that a frequency warping and a level compression had an effect on the naturalness of pitch converting speech. As results, it has been capable of defining a function between a frequency warping coefficient and a pitch conversion ratio. In 2000, we have investigated an algorithm for a speech rate control, which consisted of a local conversion and a global one temporally. And a VUS decision, which was an important factor of qualities of re-synthesized aid speech, has I been investigated also. Moreover, an evaluation system was realized on WINDOWS. As result, it has made an available of on-line evaluation tests. However, real time simulation on DSP system has not j been realized at present. In future, we are planning to device a real-time operation system based on i a board using a DSP chip, which has been purchased.