欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
为解决在说话人识别方法的矢量量化（Vector Quantization,VQ）系统中,K-均值法的码本设计很容易陷入局部最优,而且初始码本的选取对最佳码本设计影响很大的问题,将遗传算法（Genetic Algorithm,GA）与基于非参数模型的VQ相结合,得到1种VQ码本设计的GA-K算法.该算法利用GA的全局优化能力得到最优的VQ码本,避免LBG算法极易收敛于局部最优点的问题;通过GA自身参数,结合K-均值法收敛速度快的优点,搜索出训练矢量空间中全局最优的码本.实验结果表明,GA-K算法优于LBG算法,可以很好地协调收敛性和识别率之间的关系.
英文摘要：
To solve the issues that K-mean algorithm is easy to fall into a local optimal result and the design of best codebook greatly depends on selection of initial codebook in the Vector Quantization （VQ） system of speaker identification, the algorithm GA-K about codebook design is proposed by combining Genetic Algorithm （GA） with VQ based on nonparametric model. The algorithm uses the global optimiza- tion function of GA to obtain the best VQ eodebook and avoids converging local optimal result of LBG algorithm. Through the parameters of GA, the global optimal codebooks are found out in the training vectors combined with the fast convergence of K-mean algorithm. The experiments show that GA-K algorithm is more effective than LBG algorithm and it can well deal with the relations between convergence and recognition rate.
同期刊论文项目
同项目期刊论文
期刊信息
位置：