An Experiment of Sound Recognition using Machine Learning

Abstract

In this study, we conduct a machine learning experiment using the UrbanSound8k dataset to recognize urban sound It can be used in cases where sound cannot be recognized due to a variety of noises through machine learning, and in areas where hearing must be protected, such as worker safety of workers. We extract mel-spectrogram from a sound file dataset and convert it to an image format for learning and validation and then determine the accuracy of the experiment. In the future, the results of our experiment will be used with non-contact sensor information from various devices such as electromagnetic sensors to safeguard workers operating in dangerous areas. © 2020 IEEE.