基于改进深度置信网络的语音增强算法  

An Advanced Speech Enhancement Algorithm Based on Deep Belief Network

作　　者：余华[1] 唐於烽 赵力[2] Yu Hua;Tang Yufeng;Zhao Li(Jiangsu Open University,Nanjing,210065,China;School of Information Engineering,Southeast University,Nanjing,210096,China)

机构地区：[1]江苏开放大学,南京210065 [2]东南大学信息科学与工程学院,南京210096

出　　处：《数据采集与处理》2018年第5期793-800,共8页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(61673108)资助项目

摘　　要：研究了一种基于深度置信网络的语音增强算法,并针对其不足做如下改进:考虑到对应训练集中噪声种类较少,噪声特性不够丰富的情况,在频域对噪声频谱进行扰动,以丰富噪声频谱特性;考虑到不同频点的信号对系统误差的影响不一样,结合绝对听阈构造权重系数。最后选取在噪声环境下传统语音增强算法中较好的LOG-MMSE和本文改进的基于深度置信网络的语音增强算法进行了分析比较,结果证明深度置信网络的语音增强算法显示出较好性能,尤其对增强后语音质量的提升超过了LOG-MMSE方法。A speech enhancement algorithm based on deep belief network is proposed and improved for its shortcomings.Since there are few types of noise in the training set and the noise characteristics are not rich enough,the noise spectrum is disturbed in the frequency domain to enrich the noise spectrum characteristics.Considering that the signals of different frequency points have different effects on the system error,the weight coefficient is combined with the absolute hearing threshold.Finally,the better LOG minumum mean square error(LOG-MMSE)in the traditional speech enhancement algorithm and the improved deep confidence network-based speech enhancement algorithm in the noise environment are compared and analyzed.The result shows that the speech enhancement algorithm of the deep belief network exhibits excellent performance,especially the enhanced voice quality compared with the LOG-MMSE.

关 键 词：语音增强算法 深度置信网络 LOG-MMSE算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]