AN EXPLAINABLE PROXY MODEL FOR MULTILABEL AUDIO SEGMENTATION - Laboratoire Informatique de l'Université du Maine Access content directly
Conference Papers Year : 2024

AN EXPLAINABLE PROXY MODEL FOR MULTILABEL AUDIO SEGMENTATION

Abstract

Audio signal segmentation is a key task for automatic audio indexing. It consists of detecting the boundaries of class-homogeneous segments in the signal. In many applications, explainable AI is a vital process for transparency of decision-making with machine learning. In this paper, we propose an explainable multilabel segmentation model that solves speech activity (SAD), music (MD), noise (ND), and overlapped speech detection (OSD) simultaneously. This proxy uses the non-negative matrix factorization (NMF) to map the embeddings used for the segmentation to the frequency domain. Experiments conducted on two datasets show similar performances as the pre-trained black box model while strong explainable features arise. Specifically, the frequency bins used for the decision can be easily identified at both the segment level (local explanations) and global level (class prototypes).
Fichier principal
Vignette du fichier
ICASSP2024_nmf.pdf (678.15 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-04393946 , version 1 (16-01-2024)

Identifiers

  • HAL Id : hal-04393946 , version 1

Cite

Théo Mariotte, Antonio Almudévar, Marie Tahon, Alfonso Ortega. AN EXPLAINABLE PROXY MODEL FOR MULTILABEL AUDIO SEGMENTATION. International Conference on Acoustics Speech and Signal Processing, IEEE, Apr 2024, Seoul (Korea), France. ⟨hal-04393946⟩
34 View
27 Download

Share

Gmail Facebook X LinkedIn More