Research On Fusion Multi - Feature SRC Gesture Recognition Algorithm

With the rapid development of computer technology, "harmonious human-computer interaction theory and intelligence information processing" become a national strategic direction of technology development, the "virtual reality technology" and "intelligent sensing technology" has become the frontier. Gesture has been one of the most important means for human-computer interaction because of its simple and convenient expression. Past gesture recognition technology based on data glove or other intelligent wearable device has achieved stage results, but complex hardware enable the application scenarios existing many restrictions, so Vision Based Gestrue Recognition has become the mainstream. The development of Sparse Representation, Compressed Sensing theory and their outstanding performance in pattern recognition,opens up new routes of the research in the related field.But the relevant research of gesture recogniton is still very limited and slow in the level of application scenarios. Therefore,an intensive study of sparse representation-based gesture recognition algorithm has important theoretical value and practical significance.The research work of this paper mainly includes the following aspects:1、The effect of contour feature which is often used in existing gesture recognition always get a poor classification result when the number of gesture species is large. To solve this problem, the edge feature is used as the global feature of the gesture in this paper. Down sampling of the edge features will be performed, as well as a simple operation of normalization. In order to offset the effect of shape distortion and small angle rotation,improve the recognition accuracy and meet different application scenarios for different gesture rotation processing requirements,two kinds of featrues:edge feature and SURF feature will be fused to work together.By Giving different weights to the test samples before sorting, the sparse representation algorithm is used to accomplish the classification at last. Compared with existing algorithms,only one step of sparse representation classification processing is carried out,which makes the complexity of the system reduced.2、In the process of gesture segmentation, it is likely to appear the interference of face and large area of skin-like color in the scene. This paper does not need to carry out a special face and skin-like color removal program. Instead, these areas will be reserved and used as a potential gesture area. The features of them will be used as the input of the Sparse Represetation-based Classifier. Heterogeneous rejection scheme in the Sparse Representation-based Claasification theory will help to eliminate interference and get a correct results in hand gesture classification. This scheme can also avoid the independent face-subtraction steps in existing research. Thus,the complexity of the algorithm is reduced,and the performance of the recognition system is improved.In most of the existing gesture recognition scheme, only the defined gesture sample will be classified. It can not deal with the undefined gesture sample.The proposed algorithm in this paper will make use of the heterogeneous rejection scheme in Sparse Represetation-based Classifier to solve this problem.The scheme can effectively prevent the recognition errors and system misoperation caused by undefined gestures,thereby improve the system’s application value.