Performance-accuracy Optimization of Face Detection in Human Machine Interaction

In intelligence human machine interaction systems, performance and accuracy of face detector play a vital role. After building a particular system, the hardware and face detection algorithm are constants. As a result, the image resolution becomes the tool to adjust the accuracy. However, the image resolution impacts the performance of such systems. Especially, in real-time systems, detection delay can degrade the overall performance drastically. If the resolution is reduced, the system responses faster. However, the accuracy of the detection degrades. On the other hand, if the resolution increases, the accuracy increases but the performance degrades. Trial and error basis approach is an intuitive solution to gain optimum performance and accuracy. However, it is a tedious method and does not exploit all of the possibilities. In this paper, a mathematical model has been derived from experimental data using curve fitting method for performance-accuracy optimization of face detection in human machine interaction.