论文部分内容阅读
由中国科学院声学所等单位参加的“七五”重点攻关项目,“连呼语音识别和单音节识别”,已于1991年1月16日在北京通过机电部和中国科学院的鉴定验收。 连呼语音识别在语音识别当中属于难点,它的难度在于音节端点的不确定以及由于连呼引起的音变。由中国科学院声学所执行的这项工作,目标还只是小字表(100—200字)的连呼识别,识别率达80％以上,形成了一个非实时的表演系统,引进语法知识,语句的
The Key Seventh Five-Year Research Project participated by the Institute of Acoustics and other units of the Chinese Academy of Sciences, “Lien Speech Recognition and Monosyllabic Identification” was approved by the Department of Mechanical and Electrical Engineering and Chinese Academy of Sciences in January 16, 1991 in Beijing. CALL speech recognition is a difficult point in speech recognition, and its difficulty lies in the uncertainty of the syllable endpoints and the pitch change caused by the consecutive calls. Accomplished by the Acoustics of the Chinese Academy of Sciences, the goal of this work is to recognize even the small calligraphies (100-200 words) with a recognition rate of over 80%, forming a non-real-time performance system that introduces grammatical knowledge and sentences