Automatic Audio Segmentation using a Measure of Audio Novelty