在多带激励（MultiBand Excitation,MBE）语音模型的基础上提出了一种基于基音处频域误差函数特性的快速基音细搜索算法和一种基于 MBE-LPC与 VDVQ（Variable-Dimension Vector Quantization）的可分级谱幅度矢量量化方案,实现了一种 1.8～2.4 kbps可分级的 MBE语音编码器。实验结果表明木文所实现 MBE编码器在保持与常规同速率的MBE-VDVQ合成质量相当的同时能大幅度减少所需的运算量和存储空间。 Based on MultiBand Excitation (MBE) speech model, a fast pitch fine search algorithm based on the frequency domain error function at pitch and a method based on MBE-LPC and Variable Volume Dimension Vector Quantization (VDVQ) The scalable spectral amplitude vector quantization scheme implements a scalable MBE speech encoder of 1.8 to 2.4 kbps. The experimental results show that the MBE encoder implemented by Mu Wen can significantly reduce the amount of computation and storage space while maintaining the same quality as the conventional MBE-VDVQ with the same speed.