论文部分内容阅读
本文把美国政府新的标准4800bps码激励线性预测(CELP)编码器跟2400bps线性预测(LPC—10e)以及16000bps连续可变斜率增量调制△M(CVSD)的联邦标准话音编码器作了一番比较。比较包括:话音质量、安静/噪声环境下/信道中的清晰度、说话人识别、摸拟和数字实现要求、转接(tandem)编码和编码时延。业已表明,在话音质量方面,4800bps的CELP要比LPC—10e和CVSD有很大改善。虽然CELP要比CVSD或LPC—10e要求大得多的计算量,但用一个新一代的单片数字信号处理器就能实现CELP编码器。不同计算量的CELP的实现方法之间提供了互操作性(interoperability),而它的强抗扰性能可允许说话人千差万别、背景噪声环境可有多种多样,也能容忍各种不同的摸拟损伤(即幅频响应不平坦的送话器和非平坦信道)、多次转接和误码。这些特点使得CELP对于许多话音应用来说都很理想。
In this paper, the U.S. government’s new standard 4800bps Code Excited Linear Prediction (CELP) encoder is compared with the 2400bps Linear Prediction (LPC-10e) and the 16000bps Continuous Variable Squared Increment Modulation Delta M (CVSD) Federal Standard Speech Encoder Compare Comparisons include: voice quality, clarity in quiet / noisy environments / channels, speaker identification, analog and digital implementation requirements, tandem encoding, and encoding latency. It has been shown that CELP at 4800 bps is significantly improved over LPC-10e and CVSD in terms of voice quality. Although CELP requires a much larger amount of computation than the CVSD or LPC-10e, a CELP encoder can be implemented with a new generation of monolithic digital signal processors. Interoperability is provided between implementations of CELP with different computations, and its strong anti-jamming ability allows speakers to vary widely, the background noise environment can be varied, and various simulations can be tolerated Impairments (ie microphones with uneven amplitude-frequency response and non-flat channels), multiple handovers and bit errors. These features make CELP ideal for many voice applications.