Detecting Mechanical Vibrations in Televisions via Audio Spectrogram Classification

This paper presents a method for contactless detec tion of mechanical vibrations in televisions through audio spec trogram classification, utilizing Convolutional Neural Networks. The model was trained on a dataset containing simulated samples and demonstrated high accuracy, with excellent learning curves observed during training. In further evaluation with real samples the model performed well, achieving F1-Score rate of 99,02% in the test partition, confirming its potential for use in preventive maintenance processes and in addressing issues in televisions and other audio-dependent equipment, thereby enhancing the efficiency and quality of service.