Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Tuochao, Shin, D, Erdogan, Hakan, Hersek, Sinan
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2506.00273
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913868714868736
author Chen, Tuochao
Shin, D
Erdogan, Hakan
Hersek, Sinan
author_facet Chen, Tuochao
Shin, D
Erdogan, Hakan
Hersek, Sinan
contents This paper introduces SoundSculpt, a neural network designed to extract target sound fields from ambisonic recordings. SoundSculpt employs an ambisonic-in-ambisonic-out architecture and is conditioned on both spatial information (e.g., target direction obtained by pointing at an immersive video) and semantic embeddings (e.g., derived from image segmentation and captioning). Trained and evaluated on synthetic and real ambisonic mixtures, SoundSculpt demonstrates superior performance compared to various signal processing baselines. Our results further reveal that while spatial conditioning alone can be effective, the combination of spatial and semantic information is beneficial in scenarios where there are secondary sound sources spatially close to the target. Additionally, we compare two different semantic embeddings derived from a text description of the target sound using text encoders.
format Preprint
id arxiv_https___arxiv_org_abs_2506_00273
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SoundSculpt: Direction and Semantics Driven Ambisonic Target Sound Extraction
Chen, Tuochao
Shin, D
Erdogan, Hakan
Hersek, Sinan
Audio and Speech Processing
Machine Learning
Sound
This paper introduces SoundSculpt, a neural network designed to extract target sound fields from ambisonic recordings. SoundSculpt employs an ambisonic-in-ambisonic-out architecture and is conditioned on both spatial information (e.g., target direction obtained by pointing at an immersive video) and semantic embeddings (e.g., derived from image segmentation and captioning). Trained and evaluated on synthetic and real ambisonic mixtures, SoundSculpt demonstrates superior performance compared to various signal processing baselines. Our results further reveal that while spatial conditioning alone can be effective, the combination of spatial and semantic information is beneficial in scenarios where there are secondary sound sources spatially close to the target. Additionally, we compare two different semantic embeddings derived from a text description of the target sound using text encoders.
title SoundSculpt: Direction and Semantics Driven Ambisonic Target Sound Extraction
topic Audio and Speech Processing
Machine Learning
Sound
url https://arxiv.org/abs/2506.00273