本文選題：噪聲譜估計(jì) + 最小統(tǒng)計(jì)算法(MS)　； 參考：《計(jì)算機(jī)工程與應(yīng)用》2017年01期

【摘要】：噪聲譜估計(jì)算法在單通道語音增強(qiáng)方法中起著重要作用,為了改善噪聲譜估計(jì)算法對噪聲的估計(jì)和更新能力,結(jié)合最小統(tǒng)計(jì)(MS)算法,對改進(jìn)的基于控制的遞歸平均(IMCRA)噪聲譜估計(jì)算法的遞歸平均參數(shù)進(jìn)行改進(jìn),并用一階遞歸的方式對平滑功率譜的最小值進(jìn)行改進(jìn)。采用譜減法對含噪語音信號作去噪處理,從客觀和主觀兩方面對不同算法的性能進(jìn)行評價,對比分析不同噪聲不同信噪比下增強(qiáng)前后語音的分段信噪比(segSNR)、PESQ得分、MOS得分。實(shí)驗(yàn)結(jié)果表明,提出的方法能夠更好地跟蹤噪聲信號變化,改善語音質(zhì)量。
[Abstract]:Noise spectrum estimation algorithm plays an important role in single-channel speech enhancement methods. In order to improve the ability of noise spectrum estimation algorithm to estimate and update noise, the minimum statistical MSM algorithm is used to improve the performance of the noise spectrum estimation algorithm. The recursive average parameters of the improved noise spectrum estimation algorithm based on control are improved, and the minimum value of smooth power spectrum is improved by first order recursion. Spectral subtraction is used to de-noise the noisy speech signal. The performance of different algorithms is evaluated from objective and subjective aspects. The segmented SNR and PESQ scores of speech before and after enhancement are compared and analyzed in different noise and different signal-to-noise ratio (SNR). The experimental results show that the proposed method can better track the change of noise signal and improve the speech quality.
【作者單位】： 安徽大學(xué)計(jì)算機(jī)科學(xué)與技術(shù)學(xué)院;
【基金】：國家自然科學(xué)基金(No.61372137,No.61301295) 安徽省自然科學(xué)基金(No.1308085QF100)
【分類號】：TN912.35
，

本文編號：1809421