Substituting Spatial by Temporal Information in Clustering of Audio Data for Defect Diagnosis

抄録

Concrete is an omnipresent material in social infrastructures, requiring careful diagnosis for defects to ensure users' safety. The hammering test is a traditional non-destructive testing method based on sound to find defects non-visible from the surface of structures. Our previous works successfully demonstrated the possibility of discriminating defect hammering samples using fuzzy c-means with spatial information. However, spatial information can sometimes be difficult to obtain in practice. The present paper investigates the possibility to substitute spatial information by temporal information. Early results of experiments using concrete test blocks indicate that although performance is degraded, better results are obtained than using only audio information.