一种自适应速率的线谱频率参数暂时分解算法

Adaptive rate temporal decomposition algorithm for line spectral frequency parameters

为了解决语音编码中线谱频率参数经传统暂时分解算法量化后编码速率波动的问题,对修正严格的暂时分解算法进行了改进,包括定义新的事件插入法则、引入事件消除机制和基于失真门限的编码速率自适应方法;提出了一种自适应速率的线谱频率参数暂时分解算法.在新算法中,暂时分解所得失真都能受到约束控制,事件分解也更加灵活,并且能够有效抑制编码速率的波动.实验结果表明,新算法通过一定时间的自适应能够将参数编码速率稳定在所需编码速率附近,获得了较好的效果.

To solve the bit rate fluctuation problem of the line spectral frequency parameters＇ quantization in speech encoding which is caused by the conventional temporal decomposition algorithm,an adaptive rate temporal decomposition algorithm for line spectral frequency parameters＇ quantization is presented based on the modified restricted temporal decomposition algorithm,by adjusting the distortion threshold value with the new event insert rule definition and the event erasure mechanism.In the new algorithm,the distortion produced in the temporal decomposition process can be controlled and the event generation is more flexible with the effective restraint of the bit rate fluctuation.The experimental results show that,with a certain adapting time,the algorithm can adjust the parameter encoding rate to the vicinal area of the necessary coding rate stably and achieve a good performance.