汉语语音特点在多道程控电子耳蜗语音处理中的应用

The application of acoustics properties of Chinese speech sound in the Speech Processing Strategy of Multiple Programmable Cochlear Implant SPCI

为了确保电子耳蜗植入患者有较好的听觉康复效果,在研究语音处理器的编码方案时,必须紧密结合语音的声学特征。上海医科大学眼耳鼻喉科医院开发了汉语语音分析与合成软件,计算和分析了汉语拼音中元音、辅音的声学参量,并对参量合成语音的方案进行了模拟和研究。结果表明:由元音的第一或第二共振峰、辅音强频区噪音、基频等合成的语音,能较好地还原原来语音的基本声学特性。这项研究成果已应用到多道程控电子耳蜗的语音处理方案中。本文报告了1例全聋患者,在植入多道程控电子耳蜗后,能正确分辨环境声、男女声和四声,对元音与浊辅音识别正确率为85%,在段落跟踪测试中清辅音识到正确率在70%以上。本文联系语音发音机制讨论了汉语拼音中元音、辅音、四声的声学特性,并结合物理心理学提出了多道程控电子耳蜗的语音处理方案,体现了研究汉语语音特点对提高汉语语种全聋病人电子耳蜗植入效果所具有的重要意义。

To ensure good hearing habilitation for cochlear implant,it is essendal to study the acoustics properties of speech sound when make research on the speech processing strategy of multiple programmable cochlear implant SPCI,The Eye & ENT Hospital,Shanghai Medical University developed a speech sound analysis and synthesis software to compute the acoustic parameters of vowels and consonants of Chinese phonetics and to sim- ulate the speech synthesis strategy.The result suggest that the speech sound which is synthesized by the first or the second formant of vowels,the intensive frequency region of consonants and the fundamental frequency can restore the acoustic properties of the original speech sound.This result has been implied in the speech processing strategy of SPCI.One bilateral profoundly deafened person has received cochlear implant which is developed by EENT Hospital of Shanghai Medical University.This patient can discriminate environmental sound,male and female voice and four tones of Chinese phonetics successfully,the correction rate of discrimination of vowels and voiced consonants is 85 percent,and of discrimination of unvoiced consonants in speech tracking is more than 70 percent.This paper discusses the acoustics properties of vowels,consonants and four tones of Chinese phonetics with their pronunciation mechanism,and proposes the speech processing strategy of SPCI on base of the theory of psycho-physics,and reflects the essentiality of implication of theacoustics properties in the speech processing strategy of SPCI.