Angular depth map generation for gesture recognition

Depth map-based recognition for gesture and motion tracking is an efficient and economic method. What the recognition system needs is not color-image delivery but the disparity image transfer. Accuracy of depth-based tracking is cost-effectively high compared with conventional methods based on raw image processing by using the software code. This is why the disparity image is generated by the hardware without any additional software calculation. Disparity map provided by the hardware-based stereoscopic vision processing is the 8-bit gray-scale image and unnecessary pixels outside the valid region are removed in the chip. According to the angle between the object and the camera, considerable pixels are deviated compared to camera rotated with right angle orientation and this problem results in accuracy decline. This paper is devoted to implementation and evaluation of disparity-based masking, recognition and tracking including angle adjustment.