一种基于低比特率语音算法的芯片设计与分析

Design and Analysis of a Chip Based on LRC Speech Algorithm

介绍了一种低比特率语音算法,它是在MBE编码算法的基础上,利用线性预测谱代替了MBE中的傅里叶变换谱对MBE编码算法进行了改进,从而在保持语音质量的情况下使比特率更低,达到2 kbit/s。同时阐述了基于该算法的系统芯片设计,并对该算法中的基音周期参数、V/U等参数的提取进行详细的推导分析。最后对本算法与常用的码激励线性预测编码(CELP)的算法进行比较,并分析其中的原因。

A low bit rate coding （LRC） speech algorithm is introduced. This algorithm makes some improvements on the basis of MBE algorithm,and replaces the Fourier transform spectrum in MBE algorithm with linear prediction spectrum , so it can make the bit rate more lower under good speech quality and its bit rate reaches 2kbit/s. Meanwhile,chip design based on this algorithm is presented and the process of extracting parameters such as pitch, V/U judge, LPC coefficients from speech is analysed. Finally, some comparisons are made between this algorithm and conventional CELP algorithm and the reasons are investigated.