Reference Microphone Selection for Guided Source Separation based on the Normalized L-p Norm

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lohmann, Anselm, Nakatani, Tomohiro, Ikeshita, Rintaro, Delcroix, Marc, Araki, Shoko, Doclo, Simon
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917052642492416
author Lohmann, Anselm
Nakatani, Tomohiro
Ikeshita, Rintaro
Delcroix, Marc
Araki, Shoko
Doclo, Simon
author_facet Lohmann, Anselm
Nakatani, Tomohiro
Ikeshita, Rintaro
Delcroix, Marc
Araki, Shoko
Doclo, Simon
contents Guided Source Separation (GSS) is a popular front-end for distant automatic speech recognition (ASR) systems using spatially distributed microphones. When considering spatially distributed microphones, the choice of reference microphone may have a large influence on the quality of the output signal and the downstream ASR performance. In GSS-based speech enhancement, reference microphone selection is typically performed using the signal-to-noise ratio (SNR), which is optimal for noise reduction but may neglect differences in early-to-late-reverberant ratio (ELR) across microphones. In this paper, we propose two reference microphone selection methods for GSS-based speech enhancement that are based on the normalized $\ell_p$-norm, either using only the normalized $\ell_p$-norm or combining the normalized $\ell_p$-norm and the SNR to account for both differences in SNR and ELR across microphones. Experimental evaluation using a CHiME-8 distant ASR system shows that the proposed $\ell_p$-norm-based methods outperform the baseline method, reducing the macro-average word error rate.
format Preprint
id arxiv_https___arxiv_org_abs_2510_27198
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Reference Microphone Selection for Guided Source Separation based on the Normalized L-p Norm
Lohmann, Anselm
Nakatani, Tomohiro
Ikeshita, Rintaro
Delcroix, Marc
Araki, Shoko
Doclo, Simon
Audio and Speech Processing
Sound
Guided Source Separation (GSS) is a popular front-end for distant automatic speech recognition (ASR) systems using spatially distributed microphones. When considering spatially distributed microphones, the choice of reference microphone may have a large influence on the quality of the output signal and the downstream ASR performance. In GSS-based speech enhancement, reference microphone selection is typically performed using the signal-to-noise ratio (SNR), which is optimal for noise reduction but may neglect differences in early-to-late-reverberant ratio (ELR) across microphones. In this paper, we propose two reference microphone selection methods for GSS-based speech enhancement that are based on the normalized $\ell_p$-norm, either using only the normalized $\ell_p$-norm or combining the normalized $\ell_p$-norm and the SNR to account for both differences in SNR and ELR across microphones. Experimental evaluation using a CHiME-8 distant ASR system shows that the proposed $\ell_p$-norm-based methods outperform the baseline method, reducing the macro-average word error rate.
title Reference Microphone Selection for Guided Source Separation based on the Normalized L-p Norm
topic Audio and Speech Processing
Sound
url https://arxiv.org/abs/2510.27198