Abstract
Automated diagnosis systems are necessary for maintenance of superannuated social infrastructures. This paper presents an automated classification method to detect defects of materials using acoustic signals in hammering test. The approach consists of two steps. The first step is extraction of features using Short-Time Fourier Transform (STFT) and the second one is training of classifiers based on AdaBoost which is a kind of ensemble learning algorithm. We use the weak classifiers based on simple template matching method, which can consider both variable scale of amplitude and variable range of frequency. In the experiments, we discriminate between woody and metal materials by different methods of hammering test, which are tapping and rubbing. Furthermore, our method can be applied to actual diagnosis; detection of crack in plaster walls.
