Research About The Accuracy Of Interaction Based On Machine Vision In Gesture Interaction

Gesture interaction techniques based on machine vision has been gradually applied inlarge screen terminal，such as Kinect and Samsung smart TV, but they cannot provide accurateinteraction like a mouse and a keyboard which the users mostly want. There is certain gapbetween the Non-contact gesture interaction and contact equipment in operational availability,especially in operation precision. Operation precision is the main basis of the study ininteraction design, however, there are few relevant research reports in current researches.This paper mainly studied the accuracy of gesture interaction based on machine vision.We designed three different point and click experiments which were based on Fitts’ Law. Theexperiments were mainly from three aspects including the operating direction (horizontal,vertical and diagonal direction), the mapping rate equation, and the algorithm precision ofmachine vision. At the same time, we also proposed the interface design suggestion based onthe operation precision.The experimental results show that different operating direction affects the interactionaccuracy little, but it mainly affects the user’s comfort. We can get the results that thehorizontal direction is the most comfortable, the vertical direction is less, and the diagonaldirection is the worst. About the mapping rate equation, we find that, when the users dofine-tuning and coarse in the operating task, the speed should not be too fast or too slow.About the algorithm precision of machine vision, we find that when the equipment precisionis enough (better than that of the lowest interaction) and the width of the target is more thanthe minimum size, the better the algorithm precision is, the more comfortable the users feeland the more accurate the interaction is when they user gesture interaction to control.The results can be used to direct interaction design from operating direction, mappingscale equations, algorithmic precision of machine vision, and improve the availability ofgesture interaction and interaction efficiency.