Salvato in:
Dettagli Bibliografici
Autori principali: Kong, Xiangzhu, Hao, Huang, Ou, Zhijian
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2506.11630
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911224066736128
author Kong, Xiangzhu
Hao, Huang
Ou, Zhijian
author_facet Kong, Xiangzhu
Hao, Huang
Ou, Zhijian
contents This paper presents SHTNet, a lightweight spherical harmonic transform (SHT) based framework, which is designed to address cross-array generalization challenges in multi-channel automatic speech recognition (ASR) through three key innovations. First, SHT based spatial sound field decomposition converts microphone signals into geometry-invariant spherical harmonic coefficients, isolating signal processing from array geometry. Second, the Spatio-Spectral Attention Fusion Network (SSAFN) combines coordinate-aware spatial modeling, refined self-attention channel combinator, and spectral noise suppression without conventional beamforming. Third, Rand-SHT training enhances robustness through random channel selection and array geometry reconstruction. The system achieves 39.26\% average CER across heterogeneous arrays (e.g., circular, square, and binaural) on datasets including Aishell-4, Alimeeting, and XMOS, with 97.1\% fewer computations than conventional neural beamformers.
format Preprint
id arxiv_https___arxiv_org_abs_2506_11630
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lightweight and Robust Multi-Channel End-to-End Speech Recognition with Spherical Harmonic Transform
Kong, Xiangzhu
Hao, Huang
Ou, Zhijian
Audio and Speech Processing
This paper presents SHTNet, a lightweight spherical harmonic transform (SHT) based framework, which is designed to address cross-array generalization challenges in multi-channel automatic speech recognition (ASR) through three key innovations. First, SHT based spatial sound field decomposition converts microphone signals into geometry-invariant spherical harmonic coefficients, isolating signal processing from array geometry. Second, the Spatio-Spectral Attention Fusion Network (SSAFN) combines coordinate-aware spatial modeling, refined self-attention channel combinator, and spectral noise suppression without conventional beamforming. Third, Rand-SHT training enhances robustness through random channel selection and array geometry reconstruction. The system achieves 39.26\% average CER across heterogeneous arrays (e.g., circular, square, and binaural) on datasets including Aishell-4, Alimeeting, and XMOS, with 97.1\% fewer computations than conventional neural beamformers.
title Lightweight and Robust Multi-Channel End-to-End Speech Recognition with Spherical Harmonic Transform
topic Audio and Speech Processing
url https://arxiv.org/abs/2506.11630