Impact of Mixup Hyperparameter Tunning on Deep Learning-based Systems for Acoustic Scene Classification

Acoustic scene classification (ASC) refers to the identification of the environment in which audio excerpts have been recorded. It associates a semantic label to each audio recording. This task has recently drawn a lot of attention as a result of electronics such as smartphones, autonomous robots, or security systems acquiring the ability to perceive sounds. State-of-the-art sound scene classification heavily relies on deep neural network models. However, the complexity of these models makes them more prone to overfitting. The most widely used approach to overcome this concern is data augmentation. In this paper, we design and analyze the behavior of multiple deep learning-based acoustic scene classification systems. These systems are built following two deep convolutional neural network architectures which are defined with different characteristics. Moreover, this work deeply explores the use of Mixup data augmentation method and the effects of varying its hyperparameters. The obtained results indicate that proper tuning of Mixup hyperparameter significantly improves the classification performance, while considering the network architecture being employed.