期刊
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
卷 29, 期 -, 页码 1987-2000出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TASLP.2021.3082307
关键词
Radio frequency; Computer architecture; Task analysis; Neurons; Tagging; Speech processing; Feature extraction; Convolutional neural networks; receptive field regularization; acoustic scene classification; instrument detection; emotion detection
资金
- LCM-K2 Center within the framework of the Austrian COMET-K2 program
- Federal State of Upper Austria
This study demonstrates the importance of tuning the Receptive Field (RF) of Convolutional Neural Networks (CNN) for their generalization performance on various audio tasks. By controlling the RF and systematically testing different architectures, the proposed approaches significantly improve the models' generalization, surpassing complex architectures and pre-trained models. The proposed CNNs achieve state-of-the-art results in multiple audio tasks, showcasing their effectiveness in improving performance.
In this paper, we study the performance of variants of well-known Convolutional Neural Network (CNN) architectures on different audio tasks. We show that tuning the Receptive Field (RF) of CNNs is crucial to their generalization. An insufficient RF limits the CNN's ability to fit the training data. In contrast, CNNs with an excessive RF tend to over-fit the training data and fail to generalize to unseen testing data. As state-of-the-art CNN architectures - in computer vision and other domains - tend to go deeper in terms of number of layers, their RF size increases and therefore they degrade in performance in several audio classification and tagging tasks. We study well-known CNN architectures and how their building blocks affect their receptive field. We propose several systematic approaches to control the RF of CNNs and systematically test the resulting architectures on different audio classification and tagging tasks and datasets. The experiments show that regularizing the RF of CNNs using our proposed approaches can drastically improve the generalization of models, out-performing complex architectures and pre-trained models on larger datasets. The proposed CNNs achieve state-of-the-art results in multiple tasks, from acoustic scene classification to emotion and theme detection in music to instrument recognition, as demonstrated by top ranks in several pertinent challenges (DCASE, MediaEval).
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据