Application Research Of Mouse Gestures Recognition Based On Intelligent Algorithm

Mouse Gestures is a new operating mode different from shortcuts and menu operation. It refers to the use of the mouse to draw the graph and identification to complete the operation. As an efficient, innovative, convenient features, Mouse gestures is welcomed and widely used in browser. However, the application of the current mouse gesture is limited to internet browsers, and only use straight line, polyline and other simple gestures, research and application of complex mouse gestures is still relatively few.As mouse gesture recognition algorithm, BP artificial neural network is the one of the most studied. Artificial neural network is a nonlinear algorithm based on biological information processing neural network with adaptive learning, parallel processing features, and distributed information storage capacity. BP back-propagation neural network is currently the most widely used artificial neural network. BP back-propagation neural network is a multilayer feed-forward networks trained using back-propagation algorithm. BP network technology is mature, simple structure, and has been widely used in pattern recognition, function approximation and other fields.However, the standard BP neural network exist many shortcomings, for example, trapped in local minima, low learning efficiency, slow convergence etc. For overcoming these shortcomings, researchers have proposed additional momentum, adaptive learning rate, conjugate gradient method, and many improved BP neural network algorithm to improve the convergence speed and reduce errors. However, the improved neural network algorithm has not been applied to the mouse gesture recognition software.This paper studies the neural network algorithm in mouse gestures application. The paper use Matlab to simulate BP neural network improved algorithm, and then analysis the results and try to find a kind of neural network algorithm that fitting for mouse gesture recognition. As different structures of network having different capability for mouse gesture recognition, BP back-propagation neural networks with different layers, different number of neurons in the hidden layer and different transfer functions were simulated and analysis based on to mouse gesture recognition data, look for right network structure for the mouse gesture recognition. Then, based on the simulation conclusion, the paper implement a complex mouse gesture recognition software.The result shows, the mouse gesture recognition software using the network structure and improved algorithm that find out by simulation is practical, run fast,and have high recognition rate. It is considered to be a good PC desktop application tools for complex mouse gestures recognition.