基于谱减法语音增强效果研究

Study on the Effectiveness of Speech Enhancement Based on Spectral Subtraction Algorithm

文章就语音增强的发展历程以及面向藏语语音的研究现状和语音特征作了介绍和分析,并选择藏语拉萨话作为语音增强实验的语种.文章选择经典的谱减法作为语音增强方法,并介绍了谱减法原理,同时设计实验方案,实验语料使用纯净语音和噪音混合成的带噪语音,按照不同信噪比进行语音测试,得到了大量数据结果.在分析结果的基础上得出结论:使用谱减法能够明显提升语音质量,提升效果随着噪音语音相关度的多少而变化.

The origin and development of speech enhancement were introduced in this article.The speech characteristics were analyzed and the process of speech processing was introduced.Tibetan Lhasa dialect of language was chosen as experiment language.The spectral subtraction method was used in speech enhancement and the principle of spectral subtraction was also described.This experiment mixed pure speech and noise to produce noise speech with different noise and we test the noise speech in different SNR.After analysis of experiment result,we observe these phenomena that the resulted of speech enhancement decrease with improvement of SNR.A t the same time the promotion effect changes with improved relevance of noisy speech.