摘要
本文提出了声音语句输入的思想，它可以分为二个阶段处理：语音识别和语音理解，两者的接口形式为音节候选向量．通过寻求Ｘ输方向最短路径的最少元素启发式推理，必要时进行概率推理，可以取得快速、高精度的效果，再辅以良好的人机界面和机器学习，可以使声音语句输入达到实用的程度．
This paper presents a method to access computers using spoken language sentence by sentence.The matching process is divided into two steps:speech recognition and speech understanding. The former changes speech signals to some alternate syllables or phonetic letters,and the latter changes the syllables to chinese characters according to their meanings in the sentence.Parsing for fewest elements and probability reasoning are used to achieve high precision and speed. With friendly interface and matching learning,such a system can be used in real world without word limits.
出处
《计算机学报》
EI
CSCD
北大核心
1994年第2期96-103,共8页
Chinese Journal of Computers
基金
国家自然科学基金
国家教委博士学科点专项科研基金
关键词
汉字系统
声音语句输入
元音识别
Speech input,speech understanding,chinese input by sentence,fewest elements parsing
