摘要
数据挖掘是一种处理大规模的数据的方法,聚类方法是数据挖掘中常用的重要方法之一,LBG聚类方法有一些不足之处,它对初值有较强的依赖性,常常聚类到局部最优点。对该算法进行了改进,采取增加聚类个数,最后再进行合并的办法,该方法减少了漏掉可能的聚类点的机会,在实际应用中有较好的效果。将该改进后的算法应用到中国手语合成的研究中,在一定的相似范围内,得到了几乎全部中国手语规范手形,这些结果对手语合成的研究是很有意义的。
Dare Mining is the effective method for processing large amount of data, in which clustering is the common and important method. However LBG clustering still have a few shortcomings, such as, it relies on initial values and often clusters some unwanted data. Improvements have been made through adding more initial values, which reduces the chance of leaking clustering point. It is prove to be effective in Chinese Sign Language synthesis applications.
出处
《华北电力大学学报(自然科学版)》
CAS
北大核心
2001年第3期62-65,共4页
Journal of North China Electric Power University:Natural Science Edition