论文部分内容阅读
随着通信的发展,对极低速率下语音压缩编码算法的需求越来越迫切。为满足极低码率的要求并获得高质量的合成语音,提出了一种高质量的0.8kb/s声码器算法。此算法基于传统的线性预测模型,利用多帧联合的超级帧参数分模式联合矢量量化技术,以及多带混合激励,子带清浊音参数相关预测,自适应谱增强,脉冲扩散后滤波等技术。主观听觉测试显示,此声码器在0.8kb/s的速率下其合成语音不仅具有高可懂度而且具有一定的自然度,诊断押韵测试(DRT)的分数为85%,而且此声码器在10-2的随机误码的信道条件下仍然具有很好的可懂度。
With the development of communication, the demand for speech compression coding algorithm at very low speed is more and more urgent. In order to meet the requirement of very low bit rate and obtain high quality synthesized speech, a high quality 0.8kb / s vocoder algorithm is proposed. This algorithm is based on the traditional linear prediction model, the multi-frame joint super-frame parameter sub-pattern joint vector quantization technique, the multi-band hybrid excitation, the sub-band unvoiced and voiced parameters prediction, adaptive spectrum enhancement and pulse diffusion filtering. Subjective auditory tests show that the vocoder has a high degree of intelligibility and a certain degree of naturality at a rate of 0.8kb / s, and the score of the diagnostic rhyme test (DRT) is 85%. Moreover, the vocoder In the 10-2 random error channel conditions still have good intelligibility.