摘要： 介绍了一个以语音识别为目的的云南民族口音普通话语音数据库。当前，语音识别技术要走向实用必须解决用户情况多样性带来的鲁棒性问题，通常把这个问题简要地归结为“男女老幼”和“南腔北调”。作为民族文化大省的云南，共有25个少数民族，广大少数民族同胞在说普通话时明显带有地方民族口音，云南民族口音普通话语音识别研究是用户情况多样性研究的重要内容，而为之建立云南民族口音普通话语音数据库是该研究的重要基础和先决条件。

关键词: 语音数据库, 语音识别, 民族口音

Abstract: This paper describes a mandarin speech database with non-native accent of Yunnan for speech recognition. At present, the robustness of multi-user must be dealt in the progress of speech recognition from laboratory demonstration into commercial use. There are 25 nationalities in Yunnan and the mandarin spoken by the nationality people include the non-native accent and dialects of Yunnan. The research on mandarin speech recognition for speakers of the nationalities in Yunnan is very important for study of robustness of multi-user. A mandarin speech database is the base and precondition for this research.

中图分类号: 

• TN912.3