Study Of Static Hand Gesture Recognition Algorithm Based On Corner Feature Reconstruction

The most popular research direction in recent years the field of artificialintelligence for gesture recognition, because with the traditional way ofhuman-computer interaction compared, its convenient, flexible, the advantages of thehardware requirements, so that it has a very important significance in the field ofhuman computer interaction.People use traditional habits instruction input keyboard or mouse and otherelectronic equipment, control the machine to get the response then, but in somespecial environment, such as the interference of electronic devices and wetenvironment, there are strict requirements for aseptic operation lab etc. these deviceswill be restricted, and the use of man-machine touch form the so-called interaction isnot ideal. In this case, related operation through gestures of the above facing problemssolved. At the same time, bound gesture expression can from equipment, can be freeto play in a certain range, thereby freeing space constraints, to achieve the remotecontrol effect, but the distance switching requirements of hardware is also very easyto implement, such as the use of zoom lens on the target acquisition. The effect isgood or bad depends on the final gesture segmentation, feature extraction, targetrecognition algorithms.Based on the fully consult the experts in field, their literature works, simulationexperiment is carried out on a typical corner feature detection model, combined withthepopular segmentation and feature modified reconstruction algorithm based on skincolor model to study. The concrete contents include:Based on the YCbCr color spacemodel, through the experimental comparison selection,to find the parameters of themodel have strong adaptability to the light condition, toextract the targetimage segmentation gestures.Based on pulse coupled neural network, the realizationof the denoising simulation of gesture image containing noise, salt and peppernoise10%incorporation experiments,through multiple sets of simulationexperiments comparing with model parameters. Themodel of the noise reductioneffect of the best. Compared with the classical mean filtering,median filtering and other commonly used methods, it is fast, the effect is good.Research on Optimizationof gesture contour edge image using Fu Liye descriptors,makes the edge smoothingeffect can be greatly improved, so that subsequent corner detection brings greatconvenience.Based on fast corner detection CPDA of gesture image feature cornerlocation, through the simulation experiment, obtained the characteristic fingerand interphalangeal angleincluded angle point information obvious inflectionpoint, the effect is very obvious.The use of image processing technology to locateand extract the fingers, these fingersinformation aided diagonal point positioninginformation and coding. Simulation experiments show that the proposed bythinning operations can be accurate to finger position information.