频域和时域信息对连续交替取样策略汉语声调识别的贡献  

Contributions of Temporal and Spectral Cues to Chinese Tone Recognition in the Continuous Interleaved Sampling Strategy

作　　者：李春晓[1] 潘翔[1] 刘琚[2] 聂开宝[2] 

机构地区：[1]浙江大学玉泉校区信息科学与工程学院信电系,杭州310027 [2]山东大学东校区信息科学与工程学院,济南250100

出　　处：《生物医学工程学杂志》2006年第1期41-44,共4页Journal of Biomedical Engineering

基　　金：国家自然科学基金资助项目(30000041);国防973资助项目(5132103ZZT21B)

摘　　要：文献研究表明,在电子耳蜗连续交替取样策略中,时域信息特别是包络随时间的变化轨迹对汉语声调识别有重要影响。在本文中,调整原始信号的包络轨迹,使四个声调的信号具有大体相同的包络变化曲线,称为调整信号。原始信号和调整信号分别通过C IS方案模拟,进行识别率对比。结果显示,在C IS方案中,时域信息对汉语声调识别有重要影响,但是频域信息是汉语声调识别的首要因素。Recent studies have revealed temporal waveform envelope cues as a function of time having significant influence on tone recognition in continuous interleaved sampling （CIS） of cochlear implants. In this study, temporal cues of speech signal have been modulated so that to different tones have nearly the same temporal waveform envelope. The processing signal is named modulated signal. The modulated signals and original signals are processed through software emulations of cochlear-implant signal processors. The recognition score of the modulated signals and originals are compared. The result indicates that temporal cues have great influence on tone recognition, but spectral cues are the principal factor determining the identification of tones.

关 键 词：频域信息 时域信息 声调识别 基音频率 连续交替取样 

分 类 号：TH789[机械工程—仪器科学与技术]