中文摘要：

提出了一种基于声音场景分类的噪声抑制算法。算法使用调制滤波法对纯语音、纯噪音和含噪语音3种场景进行分类,并根据分类结果调整噪声抑制算法参数集,得到不同的抑制系数。本文方法在助听器测试系统中取得了良好的实验效果,场景分类正确率在95%以上。在不同噪声类型情况下,经过本文算法处理的输出语音信号取得了良好的信噪比和MOS评分的提升。本文算法可以有效地提高数字助听器输出语音质量。

英文摘要：

A new noise reduction algorithm based on acoustic scene classification is proposed. Three acoustic scenes of pure speech, noise, noisy speech are classified by modulation filter. The parameters of noise reduction algorithm are adjusted by the result of scene classification. Different attenuation coeffi- cient is adopted according to the different acoustic scene. Satisfied experimental results are achieved in the digital hearing aid testing system. Better than 95% accuracy is acquired in acoustic scene classification experiment, In the environment of different kinds of noises input, the signal-noise ratio （SNR） and MOS score are increased apparently. The quality of output speech in digital hearing aids is improved effective- ly.