摘要
针对事例检索算法中最近邻算法判断盲目、计算量大等问题,改进为聚类C-均值算法;对C-均值聚类对初值敏感,分类结果受到取定的类别数目及聚类中心初始位置的影响,及易陷于局部极小值等问题,再次将改进的算法结合改进后的最大最小距离法,以优化初始聚类,将最终改进的算法进行了仿真比较。将最终改进的算法运用于情感智能教学中,创建了面部表情的子表情模板,提高了表情的识别率。
Aiming at the shortcomings of the nearest neighbor method such as aimless searching, heavy workload etc., utilized in the case retrieval algorithms, an improved C-means clustering algorithm is presented in this paper. Also, in order to lower the sensitiveness of C-means to following three factors： initial value, classification number and initial position of the class centers, a combined algorithm of both C-means and the improved max-min distance means is given to optimize the initial cluster. At last, an application of the algorithm used in emotional intelligence teaching （especially in facial expressions）is presented in detail, creating the sub-expression templates. Simulation shows that the advanced method improves the face recognition rate.
出处
《微型机与应用》
2011年第13期72-74,共3页
Microcomputer & Its Applications
基金
广西科学研究与技术开发计划项目(桂科攻10100002-2)
关键词
CBR
事例检索
聚类C-均值算法
表情识别
CBR
case retrieval
C-means clustering algorithm
Expression recognition
