基于短时谱相估计改进型减谱法语音降噪研究  

Research on Developed Spectral Subtraction Based on Short-time General Phase Estimation

作　　者：乔彬彬[1,2] 江冰[1] 马胜 

机构地区：[1]河海大学计算机与信息学院,常州213022 [2]江苏乐众信息技术有限公司,常州213022

出　　处：《微处理机》2013年第2期43-45,共3页Microprocessors

摘　　要：针对手机等移动智能终端设备的降噪问题,通过对基本减谱法的分析研究发现其在使用中存在不足,对基本减谱法进行了功率谱修正和输入幅值谱自适应的改进,通过调节噪声系数α和功率谱系数β以提高语言处理的灵活性以及输出信号的信噪比。仿真实验对比了基本减谱法与改进减谱法对含噪语音的处理效果,结果表明改进减谱法对含噪语音的降噪效果更为明显。Aiming at noise reduction problem for intelligent terminals such as mobile phones,by researching and analyzing for the base spectral subtraction,the shortcomings are found while using.The power spectrum is corrected and input spectrum self-adaptability is improved.The noise factor α and the power spectrum coefficient β are adjusted to improve the flexibility of speech processing and the noise ratio of output signal.The simulation experiment,comparing the result of noise reduction by base spectral subtraction and developed spectral subtraction,shows that the developed spectral subtraction is better than the base spectral subtraction for noise reduction.

关 键 词：噪音特性 减谱法 观测模型 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]