Search papers, labs, and topics across Lattice.
The paper introduces SPECTRA, a novel framework for fully few-shot class-incremental audio classification that enhances performance by addressing the limitations of existing methods which rely on frozen encoders and point prototypes. SPECTRA incorporates a lightweight trainable adapter for embedding calibration, a subspace feature replay mechanism to mitigate forgetting, and a transductive optimal-transport refinement for prototype adjustment during testing. Experimental results demonstrate that SPECTRA significantly improves accuracy and reduces forgetting across three benchmarks compared to state-of-the-art approaches, validating the effectiveness of its components through ablation studies.
Forgetting is drastically reduced in audio classification tasks, with SPECTRA outperforming traditional methods by leveraging subspace structure for feature replay.
Fully few-shot class-incremental audio classification (FFCAC) requires recognizing new sound classes from only a handful of labeled examples per session, without forgetting previously learned classes and without any large base dataset. Existing methods typically freeze a pre-trained audio--language encoder and classify with point prototypes, but they suffer from significant performance degradation throughout the sessions due to generic feature representations. We propose SPECTRA, a framework built on a frozen encoder which adds three components. (i) a lightweight trainable adapter that calibrates the generic embeddings to the task; (ii) subspace feature replay, an exemplar-free anti-forgetting scheme that replays old classes by sampling from the low-rank subspace of their stored features; and (iii) a transductive optimal-transport refinement of prototypes at test time. Our central finding is that the subspace structure of the replay diminishes forgetting and outperforms naive Gaussian replay of equal variance. On three FFCAC benchmarks (NSynth-100, FSC-89, LS-100), SPECTRA improves average accuracy and reduces forgetting over current state-of-the-art methods, and our ablations statistically validate each component.
