BLINC is proposed, a training-free TTA method that remaps the predicted time-frequency mask onto a bimodal target distribution by histogram matching and improves the overall quality of both evaluated SE models on almost every target condition and matches or exceeds the loss-based TTA baselines at minimal overhead.
Abstract
Speech enhancement (SE) models degrade under domain shifts and have to adapt to unseen target domains during deployment. Most existing test-time adaptation (TTA) methods for SE do so by adapting a subset of the model weights using a self-supervised loss, which requires backpropagation at test-time and permanently alters the model. We instead recalibrate the prediction itself and propose BLINC, a training-free TTA method that remaps the predicted time-frequency mask onto a bimodal target distribution by histogram matching. At test-time, the target distribution is parameterized from blind features of the noisy recording, so neither a reference distribution from a classical algorithm nor online metric optimization is involved. BLINC improves the overall quality of both evaluated SE models on almost every target condition and matches or exceeds the loss-based TTA baselines at minimal overhead.
The main contribution is an interpretability-informed curriculum design for blind acoustic parameter estimation with clearer scope and improved reproducibility for ViT-based models.
Luís Paulo, Albuquerque Guedes, M. Petraglia et al.· 0 citations
This work frames this as Domain-Incremental Learning over acoustic environments and presents the first continual learning benchmark for multi-channel replay speech detection, evaluating a state-of-the-art beamformer-based detector over all 24 environment orderings of the ReMASC corpus with five seeds.
SpecAugment is the dominant speech augmentation policy, but researchers have studied it primarily for automatic speech recognition (ASR), and it operates on spectrograms, limiting its applicability to other speech tasks and waveform models. We propose Wav2Aug, a drop-in waveform augmentation policy that randomly sample...
Gianfranco Dumoulin Bertucci, M. Ravanelli· 0 citations
Pretrained audio spectrogram transformers are widely used as backbones of machine listening systems for sound event classification. However, they are typically trained using spectrograms with a fixed Short-Time Fourier Transform (STFT) configuration, implicitly binding the model to the time-frequency resolution seen du...
Julia Wilkins, Wei-Ching Lin, Abinaya Kumar et al.· 0 citations
This paper addresses unsupervised speech enhancement in the unpaired setting using drifting methods, where training relies on separate collections of degraded and clean audio without corresponding pairs. While recent drifting approaches enable unpaired training, they do so at a heavy cost: because the objective optimiz...
Diego Caviedes-Nozal, Liang Xu, R. Olsson et al.· 0 citations
GLaS-JEPA is presented, a framework that directly predicts the current encoder's continuous representations at masked positions, without contrastive learning, discrete targets, or separate EMA target encoders, and prevents representation collapse using SIGReg representation-space regularization.
G. Botté, Séverin Baroudi, Samir Sadok et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.