Paper ID | SPE-51.5 | ||
Paper Title | PERCEPTUAL LOSS BASED SPEECH DENOISING WITH AN ENSEMBLE OF AUDIO PATTERN RECOGNITION AND SELF-SUPERVISED MODELS | ||
Authors | Saurabh Kataria, Jesús Villalba, Najim Dehak, Johns Hopkins University, United States | ||
Session | SPE-51: Speech Enhancement 7: Single-channel Processing | ||
Location | Gather.Town | ||
Session Time: | Friday, 11 June, 13:00 - 13:45 | ||
Presentation Time: | Friday, 11 June, 13:00 - 13:45 | ||
Presentation | Poster | ||
Topic | Speech Processing: [SPE-ENHA] Speech Enhancement and Separation | ||
IEEE Xplore Open Preview | Click here to view in IEEE Xplore | ||
Abstract | Deep learning based speech denoising still suffers from the challenge of improving perceptual quality of enhanced signals. We introduce a generalized framework called Perceptual Ensemble Regularization Loss (PERL) built on the idea of perceptual losses. Perceptual loss discourages distortion to certain speech properties and we analyze it using six large-scale pre-trained models: speaker classification, acoustic model, speaker embedding, emotion classification, and two self-supervised speech encoders (PASE+, wav2vec 2.0). We first build a strong baseline (w/o PERL) using Conformer Transformer Networks on the popular enhancement benchmark called VCTK-DEMAND. Using auxiliary models one at a time, we find acoustic event and self-supervised model PASE+ to be most effective. Our best model (PERL-AE) only uses acoustic event model (utilizing AudioSet) to outperform state-of-the-art methods on major perceptual metrics. To explore if denoising can leverage full framework, we use all networks but find that our seven-loss formulation suffers from the challenges of Multi-Task Learning. Finally, we report a critical observation that state-of-the-art Multi-Task weight learning methods cannot outperform hand tuning, perhaps due to challenges of domain mismatch and weak complementarity of losses. |