Abstract:

Traditional manual methods for detecting wall hollowing suffered from strong subjectivity, low efficiency, and difficulties in large-scale application. To address these issues, this study proposed an intelligent detection method based on a fully automatic hollowing signal acquisition device and an optimized signal processing algorithm. Firstly, a fully automatic hollowing signal acquisition device capable of stable operation on building walls was designed to achieve standardized tapping and high-precision acoustic signal acquisition. Secondly, VMD and EEMD optimized by Bayesian Optimization (BO) were employed to denoise the original signals, thereby enhancing the features of hollowing signals. Then, MSC and MFCC features of the signals were extracted and fused at the frame level to form an MFCC+MSC feature set. Finally, a majority voting ensemble learning model was utilized for classification, enabling high-precision hollowing detection. The results indicate that the classification accuracy of the proposed method reaches 99.31%, significantly outperforming traditional methods. These results validate the feasibility and effectiveness of combining automated devices with optimized signal processing techniques for wall hollowing detection.

Key words: hollowing signal, acquisition device, ensemble empirical mode decomposition (EEMD), variational mode decomposition (VMD), mel-frequency cepstral coefficients (MFCC), Mel spectrum characteristics (MSC)