arXiv subjects

Yoshiki Mitsui

Publications and source records attributed to Yoshiki Mitsui.

3 recordsLinked to original sources

arXiv2026

Downstream-Task-Aware Unified Source Separation

Task-aware unified source separation (TUSS) enables a single model to handle diverse separation tasks by conditioning on input prompts. However, conventional TUSS does not account for downstream task requirements, such as whether the enhanced speech will be used for human listening or automatic speech recognition (ASR). In this paper, we propose a prompt extension framework for TUSS that incorporates downstream task information into the input prompts and switches the loss function according to the given prompt during training, enabling outputs with different signal characteristics at inference time. Specifically, we introduce an ASR-dedicated prompt paired with a regularized loss function that reduces speech artifacts to improve ASR robustness, while the standard prompt is paired with the conventional SNR loss function. Experiments on the LibriSpeech and JNAS corpora demonstrate that the proposed joint-training scheme enables a single model to improve ASR performance over noisy input across a wide range of SNR conditions by selecting the ASR-dedicated prompt, while maintaining general speech enhancement quality when the standard prompt is used.

eess.AS↗

arXiv2017

Independent Low-Rank Matrix Analysis Based on Parametric Majorization-Equalization Algorithm

In this paper, we propose a new optimization method for independent low-rank matrix analysis (ILRMA) based on a parametric majorization-equalization algorithm. ILRMA is an efficient blind source separation technique that simultaneously estimates a spatial demixing matrix (spatial model) and the power spectrograms of each estimated source (source model). In ILRMA, since both models are alternately optimized by iterative update rules, the difference in the convergence speeds between these models often results in a poor local solution. To solve this problem, we introduce a new parameter that controls the convergence speed of the source model and find the best balance between the optimizations in the spatial and source models for ILRMA.

cs.SD↗

arXiv2017

Independent Low-Rank Matrix Analysis Based on Complex Student's $t$-Distribution for Blind Audio Source Separation

In this paper, we generalize a source generative model in a state-of-the-art blind source separation (BSS), independent low-rank matrix analysis (ILRMA). ILRMA is a unified method of frequency-domain independent component analysis and nonnegative matrix factorization and can provide better performance for audio BSS tasks. To further improve the performance and stability of the separation, we introduce an isotropic complex Student's $t$-distribution as a source generative model, which includes the isotropic complex Gaussian distribution used in conventional ILRMA. Experiments are conducted using both music and speech BSS tasks, and the results show the validity of the proposed method.

cs.SD↗