To capture and track pedestrian or humans in their motion and any moving object is always a challenging task for any system. The system becomes more challenging due to the scope of variation of targets, light conditions, motion of the object. The histogram of oriented gradients (HOG) descriptor is one of the best and most popular descriptors used for pedestrian detection using Harr classifier. The HOG detector is a sliding window algorithm, which means that for any given image a window is moved across at all locations and scales and a descriptor is computed. The window is a pre trained classifier which is computed for the dataset for the descriptor. The classifier used is a linear Support Vector Machine classifier and the descriptor is based on histograms of gradient orientations. Gradient orientations and magnitude are obtained for each pixel from the pre-processed image. The dataset is created and the hit threshold is created for the descriptor for 30 frames per second for the 1000 positive images. The capture window size is reduced to 320 by 240 to get the efficiency and speed which is the limitation of the HOG.