Firstly the patterns of the multifingered hands are detailed, eight patterns are defined . the classical bayes method is used in the classification of pre-grasp of multiple fingers based on three patterns which are grasping, holding and pinching . based on the eight pre-grasp patterns, bp neural network is applied in the classification of the pre-grasp of multifingered hands and gets a good effect . the method solves the shortcoming input sample relying on the propobility density and simplified the un-insititution characters extraction . in this paper, support vector machine ( svm ) and binary-tree with clustering is applied in the classification . this method can solve the slow speed and effect with fewness sample in the classification, achieving a good effect . in this papper, we extract the characters of the regulation object with geometry characters and extact the unregulation object with the image analysis 此法解决了输入样本依赖物体的概率密度的特点,简化了分类特征提取的不直观性。本文还采用了支持向量机(svm)和聚类二叉树相结合的方法对机器人手预抓取八类模式进行分类,解决了预抓取模式分类训练速度过慢以及在分类中样本数量偏少而影响分类效果的问题,得到了较高的正确率。本文对预抓取几何形状规则的物体采用直接提取其几何特征,对于预抓取几何形状不规则的物体采用图像分析的方法进行特征提取。
However, the conditional nayve bayes method does n't consider the different features between the legitimate mai1 and the junk in the process of classifying and filtering mail and do n't take into account the loss of misclassifying legitimate mail as junk, there are some limitations on filtering mail 但是,在邮件过滤过程中,合法邮件被误判为垃圾邮件将可能给用户带来巨大的损失。传统的朴素贝叶斯算法在对邮件进行分类与过滤时,没有充分考虑到合法邮件与垃圾邮件具有的这一不同特性,因此用于邮件过滤时有一定的局限性。
The results of recognition illuminates that this model can efficiently improve the recognition rate ( and the time of recognition for only using well-bp neural network is too long, but that of combing the simple-bp neural network with bayes method is not so long for target identification when the two methods have the same recognition rate . ) 识别结果表明,这种方法能够有效地提高识别率。同样的,将bp神经网络与d-s方法相结合(bpd-s方法)来识别目标,其识别率有了进一步的提升,其稳健性相对较好,但是识别时间比bpbayes方法的要长。