噪声模型在参数音频编码算法中的应用  

Application of Noise Model in Parametric Audio Coding

作　　者：周宏[1] 陈健[1] 

机构地区：[1]上海交通大学电子工程系,上海200030

出　　处：《上海交通大学学报》2002年第12期1799-1802,共4页Journal of Shanghai Jiaotong University

摘　　要：针对现有噪声模型合成噪声偏大、与正弦波模型融合得不好的缺陷 ,提出用改进的 MP( Matching Pursuit)算法来提取残差信号中多余的正弦波 ,利用基于能量比值的停止原则 ,使残差信号尽可能地接近白噪声而迭代次数又能有效地减少 ;改进了频带边界的能量分配方式以弥补频谱分辨率的不足 .改进模型控制结构简单 ,所需比特数少且能提高合成声音的质量 .Improvements are made with respect to the drawbacks of current noise model such as annoying noise and poor fusion with sinusoid model. It adopts a modified Matching Pursuit (MP) algorithm to extract the additional sinusoids from the residual. It proposes a new stop criterion based on energy ratio, which can make the residual be as close as possible to white noise and can effectively reduce the number of iterations. It improves the energy allocation among band borders to compensate for the lack of frequency resolution. The model can improve the sound quality with simple control structure and low bit rate demand.

关 键 词：噪声模型 参数音频编码算法 停止原则 陷波滤波器 语音编码 语音信号处理 

分 类 号：TN912.32[电子电信—通信与信息系统]