【摘要】 针对加权kNN(k-Nearest Neighbor)方法在对样本进行分类时,仅仅只利用了它的k近邻点来进行分类决策的不足,提出了一种序列的加权kNN分类方法.该方法在对某个测试样本进行分类时,除了利用它k近邻点所提供的类别信息外,还有效地利用了前面已分类样本的类别信息,这使得测试样本的分类决策更加合理和有效.在Cohn-Kanade人脸库上进行的表情识别实验表明,在序列样本分类的场合,该方法的分类效果比加权kNN方法更好.更多还原

【Abstract】 Aim at the defect that weighted k-nearest neighbor method classifies one test sample only using the class information of its k-nearest samples,a sequential weighted k-nearest neighbor classification method is proposed in this paper.Not only the class information offered by k-nearest neighbor points of test sample but also the class information of previous test sample is used for classification in the proposed method.So its decision-making processing is more reasonable and effective.The experimental results of facial expression recognition in Cohn-Kanade face database show the method is better than weighted k-nearest neighbor method for the classification of sequential samples.更多还原