基于神经网络的线性预测语音编码算法

An Algorithm of the Linear Prediction Coding Based on the Neural Network

语音压缩是多媒体通信技术的重要环节 ,线性预测编码 (LPC)技术是参数编码技术的重要内容 ,线性预测是语音信号处理中最有效的方法之一。文中从LPC原理入手 ,阐述了最佳LPC系数的计算 ,针对目前自相关法和协方差等存在着估计误差的特点 ,提出一种基于神经网络的线性预测算法 ,最后通过实验数据证明这种方法既提高了解的精度 。

Speech signal compressing is an important part of multimedia communication technology, and the linear prediction coding technology is an important content of the parameter coding technique. The linear prediction coding(LPC) is one of the most important methods in speech processing system. The current auto-correlation method and the covariance method both have estimation error. Using the theory of linear prediction coding, we suggest an improved linear prediction method based on the neural network. It is demonstrated by practice that this method not only ensures system stability but also enhances the accuracy of prediction.