基于双向搜索方法的最小值控制递归平均语音增强算法
Speech enhancement approach based on minima controlled recursive averaging algorithm using bidirectional searching method
-
摘要: 语音增强效果的提高,有赖于对噪声的准确估计和对噪声变化的及时跟踪与更新。为了提高对非平稳噪声的估计和更新能力,本文基于"改进的最小值控制递归平均"(IMCRA)算法,提出了噪声谱最小值双向搜索的改进算法。该算法结合前向搜索和后向搜索谱最小值方法的特点,有效提高噪声估计的准确性、减小非平稳噪声跟踪的延迟。实验仿真表明:在非平稳噪声环境和低信噪比条件的语音信号增强处理中,本文提出的改进算法非常有效,与IMCRA算法相比,它可以获得更好的分段信噪比的提高。Abstract: A modified algorithm,in which the bidirectional searching method is used to determine the minima of the noise power spectrum,is proposed based on the improved minima controlled recursive averaging(IMCRA) algorithm. In the proposed algorithm,both minima tracking methods of forward searching and backward searching are utilized together.And the advantages of the two searching methods are combined to increase the accuracy of noise estimator and reduce the delay of the noise minima tracking.The simulated experiments indicate that in nonstationary noise environments and under lower SNR conditions,the proposed approach is very effective.In particular,compared to IMCRA,it obtains a higher improvement in the segmental SNR.