研究 GL A算法和随机松驰算法设计的线谱对码本性能以及两种算法的特点 .训练数据采用了去均值和一阶滑动平均模型预测的误差信号 ,随机松驰算法选用简化的解码器扰动算法 .传统观点认为随机松弛算法比 GL A算法每矢量索引少用 1bit.实验结果表明 ,在小训练数据量时的确如此 ,但在大数据量时它们的性能相差不大 .在线谱对的码本设计中 ,随机松驰算法设计的码本信噪比提高很小 The performance of the codebook and the characteristics of the two algorithms of the line spectrum, which are designed by the GL A algorithm and the random relaxation algorithm, are analyzed.The training data uses the error signal predicted by the first-order and the first-order moving average models, and the random relaxation algorithm uses a simplified decoder Perturbation algorithm.The traditional view is that the random relaxation algorithm uses 1bit less per vector index than the GLA algorithm.The experimental results show that when the amount of training data is small, it is true, but its performance is similar when the amount of large data is small. In the codebook design, the codebook SNR designed by the random relaxation algorithm improves little