Evaluating Similarity of Spectrogram-like Images of DC Motor Sounds by Pearson Correlation Coefficient

Three main approaches on how audio signals can be used as input to a deep learning model are: extracting hand-crafted features from signals, mapping into appropriate images such spectrogram-like ones, and using directly raw signals. Among these approaches, the usage of represents compromise regarding bias enforced by processing (seen in features) computational demands signals). When any is input, then different techniques for image become available implemented. They include assessing similarity, implementing matching, recognition. The topic this paper similarity obtained DC motor sounds. In that respect, relevant measures are first reviewed, one them - Pearson correlation coefficient applied evaluating within same class between two classes images.