Hammering Inspection System Using HPSS and Gradient Boosting with a Wall-Climbing Robot

This paper proposes an automatic judgment system that suppresses noise in the observed signal, which contains both hammering sound and noise, using Harmonic/Percussive Sound Separation (HPSS) to extract hammering sound, and then uses machine learning to judge whether the inspection area is healthy or damaged. The proposed method calculates hammering sounds from the observed signal using HPSS, and extracts features considering the timing of the hammering sound occurrence from the hammering sound. Subsequently, Gradient Boosting is used to automatically judge whether the area is healthy or damaged. The feature of the proposed method is that it suppresses noise while retaining the characteristic hammering sound necessary for judging healthy and damaged areas, and by extracting features considering the timing of the hammering sound occurrence, it allows for highly accurate classification.