Zeit-Frequenz-Verfahren für die Klassifikation transienter Signale

Abstract

The classification of transient signals is a task which is of great significance for industry and research. In general, time-frequency methods are used for feature extraction. In this paper classic techniques and orthogonal wavelet techniques are used alternatively for feature generation. Using touched piano chords a classifier checks, whether the corresponding piano can be identified and therefore allows a comparison of the techniques used. Classification results and examples of visualizations, which show the features in the time-frequency plane, are discussed.