论文部分内容阅读
详细分析了小波变换、镜像滤波器组和散射矩阵之间的关系，提出了由散射矩阵的基本因子来构造镜像滤波器对，得出了具有任意消失矩的紧支正交小波的参量化法及基于正则性约束的参量优化．给出了一种改进的码激励线性预测（ＣＥＬＰ）编码方案，该方案采用由参量化法求解的具有最大消失矩的紧支正交小波和ＣＥＬＰ算法，可保证在４．８ｋｂｉｔ／ｓ的码率下得到高质量的语音．使用小波变换替代了传统的ＣＥＬＰ算法中随机码本的搜寻，在保证语声质量的基础上大大降低了运算量和存储量，算法简单，易于实时实现．
The relationship between wavelet transform, image filter bank and scattering matrix is analyzed in detail. It is proposed that the mirror filter pair is constructed by the basic factors of the scattering matrix, and the parametric method of the compact orthogonal wavelet with arbitrary vanishing moments And parameter optimization based on regularity constraints. An improved Code Excited Linear Prediction (CELP) coding scheme is proposed. This scheme adopts the clamped orthogonal wavelet with maximum vanishing moments and CELP algorithm, which is solved by parametric method, Get high-quality voice rate. The wavelet transform is used to replace the random codebook search in the traditional CELP algorithm, which greatly reduces the computational complexity and storage on the basis of ensuring the speech quality. The algorithm is simple and easy to implement in real time.