DYNAMIC AND STATIC HAND GESTURE RECOGNITION SYSTEM FOR HUMAN COMPUTER INTERACTION USING LOW END WEBCAMERA IMAGE ANALYSIS

摘要

One of the most important aspects of future intelligent and interactive computing systems involves an efficient human -computer interaction. Gesture recognition is one way to achieve it. It is the process by which the gestures made by the user are recog nized by the receiver. Primitive text user interfaces have become obsolete and even Graphical User Interface (GUI) is mainly limited to the mouse and the keyboard. Gesture recognition can be used instead, to enable the system to analyse user requests without any physical device. Gesture recognition can be done by using special gloves but they are cumbersome to use. Instead of this, a more user-friendly approach is by using a Vision-Based system. Gesture recognition can be conducted with techniques from computer vision and image processing. There are three important phases, namely, hand recognition and tracking, symbol recognition, and lastly, invoking an application. The first phase involves setting up a fixed background and subtracting the same from the next frame in order to find contour of the hand. Once the contours have been found, the gesture traced by the user is recorded and it is checked against the existing database of symbols. Finally, once the gesture has been matched, the application corresponding to the gesture is in voked. The gesture recognition system recognises nine symbols. Each symbol was tested twenty times to determine the recognition rate and accuracy of the system. The developed system recognises symbols at a good accuracy rate of 81.11%. The time taken to re cognise symbols is around 0.035 seconds on average. The symbol B has the highest accuracy rate of 90% and it is also recognised the fastest with an average time of 0.02 seconds. The time taken to invoke applications once the symbol is successfully recognis ed is negligible. While some approaches have higher accuracy, they are also slower by over 250 milliseconds and occupy more thrice as much spac e in memory.