Human detection plays an important role in security surveillance and computer vision. The process of human detection is very complex due to variant feature of human such as color, texture and shape and size. The process of feature extraction imparts a major role in human detection technique. Now a days used classification technique to define the feature of human. The classification process define the pattern of feature for the process of detection, the process of features generates a bag of feature for the process of classification technique. In this paper improved the support vector machine classification technique for the classification of human detection. The improved support vector machine is called cascaded support vector machine. The cascading of support vector machine improved the process of human detection. Our proposed algorithm implemented in MATLAB 7.8.0 software and used human video of different location. Our empirical evaluation of experimental result shows that the proposed methods give a better result in compression of support vector machine classifier.