摘要
本文论述了基于神经网络模型的非特定人汉语语音识别。我们采用24人(12人用于训练,12人用于测试)的语音数据对汉语十个数字和十个孤立字进行了实验,取得了96.3％(10个数字)和97.2％(十个汉字)的识别率。
Chinese voice recognition of speaker-independent in a neural network model is described in this paper. The Voice data of 10 numbers and 10 Chinese isolated words by 24 speakers (12 used for trainins, 12 used for testing) were tested. Discriminability of 96.3％ (for 10 numbers) and 97.2％ (for 10 Chinese words) were obtained.
出处
《电声技术》
北大核心
1992年第9期2-6,共5页
Audio Engineering
关键词
语音识别
神经网络
汉语
voice recognition neural network speaker-independent
