Applying machine learning for hand posture recognition from a series of depth images

Image recognition one of the most difficult issue, is becoming an applied research in recent times. The possibility allowing computers to recognize human’s hand posture real-time gradually popular and involved many fields. It also integrated into so products. Thus, this study proposes approach for faster more accurate image recognition. With development technology, camera device has had significant changes; can capture both RGB images depth. This makes identification low-light conditions accurate, which improves performance. Random Forest algorithm evaluated be efficient, faster, than previous algorithms used model testing. We collected about 10,000 data samples, then trained with nearly 80% accuracy results. In addition, presents some results gesture human hands using information obtained from Kinect sensor; ability train quite large speed faster. Research control applications serving life industrial production.