【摘要】 尽管码激励线性预测编码(CELP)算法在中、低比特速率中已提供了较好的语声质量,但对随机码本的全搜索,其计算是极其复杂的。为此研究了一种用繁殖码本作为激励源的有效搜索方法,同时讨论了决定最小均方误差激励码字的全搜索过程。通过用繁殖码本,使CELP编码的计算量减少近一个数量级。更多还原

【Abstract】 Although CELP (Code-Excited Linear Predictive) has provided good quality synthetic speech at medium and low bit rates, yet the computation of exhaustive search for stochastic code book is extremely complex. So an efficient search method by using populating code book are developed, and search procedures for determining the minimum mean-square error excitation code word are discussed. The computational cost of CELP coding can be reduced more than an order of magnitude by using the populating code book.更多还原