Classifying disc defects in optical disc drives by using time-series clustering

Optical disc drives are subject to various disturbancesand faults. A special type of fault is the so-calleddisc defect. In this paper we present an approach for discdefect classification. It is based on hierarchical clustering ofmeasured signals that are affected by disc defects. The timeseriesare mapped into a feature space after which the featurevectors are clustered in a hierarchical fashion. Finally, signalsare fitted onto the clusters to obtain single representationsfor each fault class. The resulting class descriptions can thenbe used for (on-line) classification of new disc defects. Theapproach is evaluated by applying it to a set of test data.