基于谱减的Robust电话语音识别方法 |
| |
引用本文: | 韩纪庆,秦兵,王承发.基于谱减的Robust电话语音识别方法[J].电声技术,2000(12):35-37. |
| |
作者姓名: | 韩纪庆 秦兵 王承发 |
| |
作者单位: | 哈尔滨工业大学计算机科学与工程系, 黑龙江哈尔滨 150001 |
| |
基金项目: | 国家863计划基金!(863-306-ZT03-02-3),深圳华为公司科技基金 |
| |
摘 要: | 谱减技术是语音增强的有效方法,并被成功地应用到了车辆等环境下的Robust语音识别中。文章对将谱减技术应用到电话语音识别中的加性噪声去除上进行了尝试。实验验证该方法有利于改进电话语音识别的性能,对不特定话音84个孤立词的实验,误识率下降了21.8%。
|
关 键 词: | 谱减 加性噪声 Robust语音识别 电话 |
修稿时间: | 2000年10月25日 |
Robust Telephone Speech Recognition Based on Spectral Subtraction |
| |
Abstract: | Spectral subtraction is an effective method of speech enhancement and has been successfully used for robust speech recognition in vehicle environment. In this papar the spectral subtraction is used to eliminate the additive noise for robust telephone speech recognition. Experiments show that the method is helpful to improve the performance of telephone speech recognitionand a 21.8% word error reduction is obtained under testing on a speaker independent 84 isolated word task. |
| |
Keywords: | spectral subtraction additive noise robust speech recognition |
本文献已被 CNKI 维普 万方数据 等数据库收录! |