A Highly Robust Audio Fingerprinting System