ABSTRACT
This paper proposes a novel approach for recogniz-
ing chords played on a guitar by using stereo cameras.
This process is done in real time by recognizing the
patterns of ¯ngers' pushing positions detected from
input images. The program detects the position of a
player's ¯ngers by using on-line adaptation of color
probabilities and a Bayesian classi¯er. Therefore the
system is able to cope easily with considerable illumi-
nation changes and a dynamic background. ARTag
is utilized to calculate the extrinsic parameters as an
online process, so that the system can recognize the
correct chord, even if the guitar is moving. By uti-
lizing a triangulation method on stereo cameras, the
3D positions of ¯ngers are recognized when a guitar
string is pressed. It used the advantages found in
PCA (Principal Component Analysis) to e±ciently
deal with complex chord classi¯cation. The experi-
mental results have revealed that the proposed sys-
tem is e®ective in real time also.