Figure 1

Schematic diagram of the signal processing and classification system. The AE signals were preprocessed with a median filter. In the following step they are grouped with a hierarchical clustering procedure. An averaging step was implemented in each cluster to improve the SNR. This is followed by a feature extraction procedure in time and frequency domains. On the test data, the feature extraction and classification steps were executed when the signal envelope exceeded a predefined threshold. The TOA is calculated by detecting the P-waves with an SVM classifier.