摘要
电子耳蜗现有产品所采用的语音处理算法以传递语音信号的幅度信息为基础,丢失了声调信息,因此造成我国电子耳蜗植入者对声调的识别还存在困难。声调信息主要蕴涵在语音基频中,在电刺激下,为了提高基频检测的准确性进而提高对声调的识别率,并对刺激速率进行准确选择,在前期提出的9度速率电刺激编码策略的基础上,通过自相关算法及平均幅度差函数的联合使用来提高基频提取的准确性,同时采用动态步长搜索解决分类饱和的问题,从而实现对刺激速率的准确选择。结果显示本研究提高了基频提取的准确性和对刺激速率进行动态选择的准确性。
Present cochlear implant products are based on the algorithms that mainly transmit the amplitude informa- tion of the speech signal, while the tonal information is lost. Thus it is difficult for the Chinese implant carrier to identify tonal information of mandarin. Tonal information is implicated in fundamental frequency. In electric-stimu- lus, in order to improve the detection accuracy of fundamental frequency and select the stimulation rate accurately, an algorithm that combines the autocorrelation function and average magnitude difference is proposed in this paper. At the same time, a dynamic step searching method is introduced to solve the saturation problem and accurately clas- sify the levels of electric-stimulus rate based on the 9-level electric-stimulus encoding algorithm we proposed earlier. Study results show that the detection accuracy of fundamental frequency is improved and the selection of electric- stimulus rate becomes more accurate.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2012年第2期271-278,共8页
Chinese Journal of Scientific Instrument
基金
国家"十一五"科技支撑计划(2008BAI50B08)
国家自然科学基金(60871083)
北京市自然科学基金(3082012)资助项目
关键词
电子耳蜗
声调
基频提取
动态分类
cochlear implant
tone
fundamental frequency detection
dynamic classification
