Depth-based Fingertip Detection for Human-Projector Interaction on Tabletop Surfaces

Abstract

This paper presents a method for detecting fingertips using depth sensors for human-projector interaction on tabletop surfaces. We propose a complementary fingertip model consisting of 3D planar points on both a finger and background. Based on our model, we propose a cost function which enforces finger model points to be fitted on a detected human body region and enforces background model points to be projected on detected background surfaces. The fingertips are then detected by checking the final costs of the models fitted around several feature points. The experimental results show that the proposed method detects fingertips accurately with respect to various angles of depth sensors.