基于支持向量机的人类活动识别，以日常生活中的10个活动进行识别。-Support Vector Machine (SVM) was first proposed in

1995 by Cortes and Vapnik [15] for solving classification and

regression problems. The solving strategy of SVM on the

multiple classification problems is commonly “one to one”

strategy, whose basic thought is that the classification problems

with N categories will be decomposed into　()1/2 NN−　 binary

classification problems to deal with, and meantime,

()1/2 NN−　training classifiers needed to be trained. In the

training process, any two categories of all the N categories

would be selected as one group. Assume that we have five

categories, and these categories are labeled ,,, , ABC DE.

Figure 3 shows the

() 551/2 ×−　classifiers according to “one to

one” tragedy.