基于改进最小值搜索的IMCRA噪声估计算法

IMCRA noise estimation algorithm based on improved minimum search method

当噪声水平升高时,现有的噪声估计算法存在跟踪时延和估计不准确的问题,为提高噪声估计的准确性,对改进的最小值控制的递归平均噪声估计算法(improved minima controlled recursive averaging,IMCRA)中的最小值搜索方法进行改进,利用连续最小值跟踪算法取代最小值统计算法,打破求解最小值受窗长影响的现状,减少跟踪时延;提出一种基于语音存在概率的偏差补偿函数模型,偏差补偿的大小由各个频带决定。实验结果表明,不管是平稳还是非平稳噪声环境,改进后的算法都能有效提高增强后语音的质量。

When the noise level increases, the existing noise estimation algorithm has the problem of tracking delay and estimation inaccuracy.To improve the accuracy of noise estimation,the minimum value search method in IMCRA was improved.The continuous minimum tracking algorithm was used to replace the minimum statistical algorithm,breaking the current situation of the minimum value affected by the window length and reducing the tracking delay. A bias compensation based on the existence probability of speech was proposed.In the function model, the magnitude of the offset compensation was determined by each frequency band. The experimental results show that the improved algorithm can effectively improve the quality of the enhanced speech, whether it is a stationary or non-stationary noise environment.