改进微粒群算法实现人脸表情准确识别

Improved PSO algorithm to achieve accurate facial expression recognition

将用正弦函数描述的微粒群算法引入到模糊c均值聚类算法中,增加了模糊c均值聚类算法隶属度矩阵初始化的多样性,减慢了局部搜索,促进在全局范围内的寻优搜索,以有效克服模糊c均值聚类算法过分依赖初始值和容易陷入局部极小值的缺点,并且将其用于人脸表情分类,跟其他人脸表情分类方法相比较。实验证明,本文的新算法有速度快,准确率高的优点。

Nonlinear inertia weight particle swarm optimization（NWPSO） that discribed with sin function is introduced in the fuzzy c-means clustering algorithm（FCM） to increase the initialization multiplicity of the degree of membership matrix.The algorithm reduces speed of the partial search,and optimization search in the overall situation scope is promoted,so the metod overcomes the the shortcoming that fuzzy c-means clustering algorithm relys on the starting value excessively and easy to fall into the partial minimum.Then the algorithm is applied in the field of facial expression classification,also compares it with other facieal expression taxonomic approach。 The experiments prove that the new algorithm has the merits that speed be quick,and rate of accuracy be good.