Adaptive time-frequency feature resolution network for acoustic scene classification. (30th June 2022)

• Record Type: Journal Article
• Authors: Zhang, Tao; Liang, Jinhua; Feng, Guoqing
• Is Part Of: Applied acoustics. Volume 195(2022)
• Publication Date: 2022-06-30
• Subjects: Acoustic scene classification -- Convolutional neural network -- Adaptive time-frequency feature resolution network; Acoustical engineering -- Periodicals; Periodicals; 620.2
• DOI: 10.1016/j.apacoust.2022.108819
• Languages: English
• ISSNs: 0003-682X

Abstract

Highlights: A priori of related sound events is imposed for acoustic scene classification. Self-attention mechanism is used to automatically select the optimized feature. Time-frequency feature resolution is utilized to design audio-related CNN. BiFPN is adopted to transmit information across multiple resolutions. Normalized mixup is used to generalize the proposed model by data augmentation. Abstract: Convolutional neural networks with time-frequency feature representation for acoustic scene classification have been attracting increasing attentions. However, most of the existing methods are restricted to design a plain discriminative model with low-level feature representations and hard to leverage a priori knowledge of sound events in a particular scene. This provides some leeway to boost the performance. In this paper, Adaptive time-frequency Feature Resolution Network (AdaFRN) is proposed to embrace the progress in attention mechanism, feature fusion, and very deep architecture. Specifically, a multi-resolution attention distributor is applied to automatically select a proper feature resolution to capture remarkable events and a bi-direction feature pyramid network is utilized to transmit information across multiple scales. Extensive experiments demonstrate that the proposed model yields better results than state-of-the-art methods. With the normalized mixup strategy, our model outperforms the state-of-the-art method by 1.6%.