CP-JKU Submissions to Dcase’20: Low-Complexity Cross-Device Acoustic Scene Classification with RF-Regularized CNNs

This technical report describes the CP-JKU team’s submission for Task 1–Subtask A (Acoustic Scene Classification with Multiple Devices) and Subtask B (Low-Complexity Acoustic Scene Classification) of the DCASE-2020 challenge [1]. For Subtask 1. A, we provide our Receptive Field (RF) regularized CNN model as a baseline, and additionally explore the use of two different domain adaptation objectives in the form of the Maximum Mean Discrepancy (MMD) and the Sliced Wasserstein Distance (SWD). For Subtask 1. B, we investigate different parameter reduction methods such as Pruning, while maintaining the receptive field of the networks. Additionally, we incorporate a decomposed convolutional layer that reduces the number of non-zero parameters in our models while only slightly decreasing the accuracy, compared to the full-parameter baseline.