Research And Application Of Continuous Gesture Segmentation And Recognition Method

On-line gesture recognition and segmentation is widely used in games,sign lan-guage recognition and human-computer interaction,etc.The related research work has been popular topics in computer vision.In recent years,with the rapid development of depth camera equipment,more and more good research results about gesture recogni-tion have appeared.Now,the research of gesture segmentation algorithm has achieved good results,but there are some unsolved problems.So we propose a real-time algorithm that calcu-lates the probability of gesture label distribution by frame and then divides successive gestures.Besides,we put forward a novel feature extraction scheme on depth video.In each depth image,we take several random points from the motion region and select a group of relevant points for each random point.Depth difference between each random point and its relevant points is calculated.The results are used to generate the random point's feature.Then we work out the probability distribution vector(PDV)for each frame in the video.Experimental results show our method is competitive to the state-of-the-art methods.Then,we found that there are some shortcomings in labeling gestures from some datasets.So we put forward a idea that tagging gesture sequences more granularly.Be-sides,we present a feature extraction method based on skeleton and depth information.And LSTM model is used for recognition.Then,we record the data set for experi-ment.The experimental results demonstrate the value of our ideas and the effectiveness of our algorithm.Third,we apply the gesture recognition algorithm to the virtual try-on system.We design three static gestures?four dynamic gestures and a real-time gesture recognition algorithm.And our scheme achieves good results.