论文部分内容阅读
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。
In order to solve this problem, a speech noise prediction algorithm based on speech linear prediction is rapidly deteriorated under noise environment. An improved noise robust sparse linear prediction algorithm is proposed. Firstly, Student’s t-distribution is used to model sparse linear predictive residuals, and the effects of additive noise are explicitly considered to improve the robustness of the model, so as to construct a complete probabilistic model. Then, the approximate posterior distribution of the model parameters is deduced by using the variant Bayesian method, and finally the noise robust sparse linear prediction parameter estimation is achieved. 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 proposed algorithm has advantages over many indicators and has better robustness to environmental noise and spectral distortion The degree of smaller, which can effectively improve the noise environment of voice quality.