论文部分内容阅读
人们面对面交谈与通过电话讲话所听到的话音之间有很大差异。典型的例子就是无线电调频广播中，听众通过电话与电台节目主持人的对话。这种音质上的差异要求我们研制从普通电话话音中产生宽带话音的算法。此算法有两个特点：一个是基于代码簿映像的频谱包络重建；另一个是采用以重建谱包络表示话音波形的话音信号重建。由于此算法无需任何附加的传输信息就能产生高质量话音，所以它在话音编码领域中十分重要且有用。以声距标准和收听测试来评估，证实此算法性能良好。
There is a big difference between what people hear face-to-face and what they hear when speaking over the phone. A typical example is a radio-FM radio in which listeners talk to radio show hosts over the phone. This difference in sound quality requires us to develop algorithms that generate broadband speech from normal telephone voice. The algorithm has two characteristics: one is spectral envelope reconstruction based on codebook image; the other is reconstruction of speech signal by using spectral envelope to reconstruct speech waveform. Since this algorithm produces high-quality speech without any additional transmission of information, it is very important and useful in the field of speech coding. Evaluated by the acoustic distance standard and listening test, this algorithm is proved to be performing well.