Audio Mamba: Selective State Spaces for Self-Supervised Audio Representations
This work addresses the need for efficient and effective general-purpose audio representations for audio recognition tasks, though it is incremental as it adapts an existing method from language modeling to audio.
The authors tackled the problem of learning self-supervised audio representations by proposing Audio Mamba, a selective state space model, and found that it consistently outperformed comparable self-supervised audio spectrogram transformer baselines on ten diverse audio recognition tasks.
Despite its widespread adoption as the prominent neural architecture, the Transformer has spurred several independent lines of work to address its limitations. One such approach is selective state space models, which have demonstrated promising results for language modelling. However, their feasibility for learning self-supervised, general-purpose audio representations is yet to be investigated. This work proposes Audio Mamba, a selective state space model for learning general-purpose audio representations from randomly masked spectrogram patches through self-supervision. Empirical results on ten diverse audio recognition downstream tasks show that the proposed models, pretrained on the AudioSet dataset, consistently outperform comparable self-supervised audio spectrogram transformer (SSAST) baselines by a considerable margin and demonstrate better performance in dataset size, sequence length and model size comparisons.