BDO与VMD-EAM算法融合的单通道语音增强模型  

Single-channel Speech Enhancement Model Based on BDO and VMD-EAM Algorithms

作　　者：王洪涛 毛露露[2] WANG Hongtao;MAO Lulu(School of Foreign Studies,Xianyang Normal University,Xianyang 712000,China;Experimental Training Center,Xianyang Normal University,Xianyang 712000,China)

机构地区：[1]咸阳师范学院外国语学院,咸阳712000 [2]咸阳师范学院实验实训中心,咸阳712000

出　　处：《自动化与仪表》2024年第9期131-137,共7页Automation & Instrumentation

基　　金：咸阳师范学院科研计划重点项目(XSYK21030)。

摘　　要：为了改善VMD在单通道语音增强中关键参数选择难、声学特征缺失多的问题,该文提出了一种BDO与VMD-EAM算法融合的单通道语音增强模型。利用DBO对输入的含噪音频进行处理,得到分解模态数与惩罚因子的最佳组合,实现VMD在分解中关键参数的自适应寻优,再根据EAM的相似度评估结果完成IMF分量的分类,对噪声分量的高频信息使用小波阈值法进行滤除,最后基于IVMD-EAM模型的重构原理实现信号重构。仿真验证及实测实验表明,所构建的语音增强模型可有效抑制噪声并提升英文语音的增强效果,各项性能评价指标也显著优于其他2种单通道语音增强模型。In order to improve the problem that it is difficult to select the key parameters of VMD and the lack of acoustic features in single-channel speech enhancement,a single-channel speech enhancement model based on BDO and VMD-EAM algorithm is proposed.Firstly,the DBO is used to process the input noisy speech to obtain the optimal combination of the number of decomposition modes and penalty factor,so that the adaptive optimization of the key parameters of VMD is realized.Then,the classification of IMF component is completed according to the similarity evaluation results of EAM,and the high-frequency information of noise component is filtered by wavelet threshold method.Finally,the signal was reconstructed based on the reconstruction principle of IVMD-EAM model.Simulation and real experiments show that the proposed speech enhancement model can effectively suppress noise and improve the enhancement effect of English speech,and all performance evaluation indicators are significantly better than the other two single-channel speech enhancement models.

关 键 词：语音增强 单通道 BDO VMD-EAM算法 信号重构 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]