|
|
| Introduction |
| Mouse Gesture Recognizer is a program for analysing mouse movement patterns using a neural network. The recognition of mouse gestures is used for intuitive user interfaces and became popular with the Opera web browser. Because the recorded patterns are quite imprecise, an algorithm with enough tolerance and the ability to learn is required. Those requirements resulted in the choice of a feedforward neural network with a backpropagation learning algorithm. The program is written in C# and targets the .NET platform. At the moment it only works on Windows systems. Mouse Gesture Recognizer was written by Robert Hahn. A full documentation is available in german.
|
| Requirements |
|
| Download |
| Functionality |
| The user draws directly on the form while pressing the right mouse button. During the movement a list of points is recorded, which is normalized to a fixed number of anchor points later. Two sequent points form a line whose orientation in the coordinate system (sinus and consinus component of the angle) is used as input value of the neural network. The neural network consists of 3 layers. The input layer has a neuron for each anchor point of the gestures. The hidden layer has a configurable number of neurons with a default of the half number of input neurons. Finally the output layer has a neuron for each gesture that can be detected. Example: A project to detect 5 different gestures, lets say Left, Right, Up, Down and Circle whose patterns each have 15 anchor points would require 28 input neurons (15 points result in 14 lines each with a sinus and a cosinus part) and 5 output neurons. The configuration is done using the property grid on the right side. Make sure you expand all nodes to see the available options. The configuration, including the patterns and the trained network, can be saved to a file. Additionally it is possible to export and import gesture patterns only. The source code is seperated in 3 parts, a generic neural network library, a library for the gesture recording and recognition and finally the demo application providing the GUI.
|
| Installation |
| Before installing the application, the .NET framework should be installed. Afterwards the installation can be started by running Setup.exe or download the source code and compile it yourself. To deinstall just delete the created folder or use the Add/Remove Software option of the control panel. The installer is not copying any files to system folders or creating any registry keys.
|
| Screenshots |
|
|
Copyright © Robert Hahn