to damonstrate the generality of our approach, We have further evaluated our method on a multi-view IXMA action dataset following the same procedure used fou the KTH and Weizmann dataset. On the IXMAS dataset,we have tested our method using the top 200 boosted features,not only on each data from all five cameras. the overall recognition accurcy from applying multi-view fusion is 92.3% which significantly exceeds all the other results listed in Table 8.for each single view, our method achieves the best results among all the methods. The NBNN classifier outperforms the linear SVM,3nn and Random forest clssifiers and improves on the accuracies for both single and multiple views.