改进的噪声鲁棒语音稀疏线性预测算法

发布时间：2018-04-05 08:13

本文选题：线性预测　切入点：算法收敛　出处：《声学学报》2014年05期

【摘要】：语音线性预测分析算法在噪声环境下性能会急剧恶化,针对这一问题,提出一种改进的噪声鲁棒稀疏线性预测算法。首先采用学生t分布对具有稀疏性的语音线性预测残差建模,并显式考虑加性噪声的影响以提高模型鲁棒性,从而构建完整的概率模型。然后采用变分贝叶斯方法推导模型参数的近似后验分布,最终实现噪声鲁棒的稀疏线性预测参数估计。实验结果表明,与传统算法以及近几年提出的基于l_1范数优化的稀疏线性预测算法相比,该算法在多项指标上具有优势,对环境噪声具有更好的鲁棒性,并且谱失真度更小,因而能够有效提高噪声环境下的语音质量。
[Abstract]:The performance of speech linear predictive analysis algorithm will deteriorate rapidly in noise environment. To solve this problem, an improved robust sparse linear prediction algorithm is proposed.First, the student t distribution is used to model the sparse linear prediction residual of speech, and the effect of additive noise is explicitly considered to improve the robustness of the model and to construct a complete probabilistic model.Then the variational Bayesian method is used to derive the approximate posterior distribution of the model parameters, and finally the robust sparse linear prediction parameter estimation of noise is realized.The experimental results show that compared with the traditional algorithm and the sparse linear prediction algorithm based on l-1 norm optimization proposed in recent years, the algorithm has many advantages, such as better robustness to environmental noise, and less spectral distortion.Therefore, the speech quality in noisy environment can be improved effectively.
【作者单位】：解放军理工大学;
【基金】：江苏省自然科学基金(BK2012510) 国家博士后科研基金(20090461424)资助
【分类号】：TN912.3

【相似文献】