Research Of Method&Application On Robustness-Oriented Gesture Recognition

Human computer interaction is one of the most significant technologies in the21stcentury, which gradually progress from the computer centered phrase to the human-oriented.Besides, the method of interaction steady developed to the phrase of adapt to the humanbehavior. Gesture is an intuitive, natural, easy way of human-computer interaction, whichenables the natural hand interaction based on computer vision as an important part.Gesture acted as one of the most important input methods in the modernhuman-computer interaction development, which input the information of hand action intocomputer. Meanwhile, computer respond the distinguished action according to the differenthand action through the process of algorithm applied in the software, which implement thecommunication and information exchange between human and computer.In this thesis, the author mainly discussed the following aspects based on his research ofgestures segmentation and recognition under complex background.1. Camera calibration based on Particle Swarm Optimization algorithm.Practicing Particle Swarm Optimization algorithm into camera calibration, priorknowledge about the model was not required, predetermined camera specific parameters onlyusing Particle Swarm Optimization algorithm to the approximation of complex nonlinearmapping between powerful ability to implicitly camera calibration, for traditional cameracalibration technique gives a new idea. Calibration block calibration using cube boxes as incamera calibration, making simple, experimental low requirements. Experimental resultsshow that high calibration accuracy, reliability, and the algorithm need not calculate theJacobian matrix, compared with the traditional optimization algorithm has the advantage ofsimple and practical.2. Template based Manual edge thinning algorithm.In order to staff organized into single edges, on the basis of image segmentation,improvement of OPTA algorithm(One-Pass Thinning Algorithm), improved retention template,used to check the manpower edge. Experiments showed that the algorithm refinement of highprecision, good reliability, but the time complexity is slightly higher than the OPTA algorithm,due to the method by adding calculated to achieve the objective of high reliability, real-time systems require further optimization.3. Hand Gesture recongnition algorithm based on Local InformationIn the context of complex, using skin color information cannot only split out staff, thisblend of colour and gesture shape information, can be better identified from complexbackgrounds. First, establish a gesture library, as a template for gesture recognition library.When the hand gesture segmentation, removed and the color of the skin color is not close,retain the color of a similar colour and colour, such as face color, and so on, then an imageedge detection, shape as a gesture of the test results information, and then compared withgestures in the template gallery. Experiments show that manpower to better recognize themethod information, and high reliability.A new idea is presented in this thesis. Compared with some other algorithms in theexperiments, the author verifies the validity of the algorithm.