Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.29164

XingChen-UniFE: A Unified Generative Speech Front-End via Complementary Acoustic-Semantic Modeling

Kelan Kuang·Qihan Hu·Jian Kang·Jie Li·Yongxiang Li·Yi Liu
View original source ↗Read source document ↗Export metadata ↓

Abstract

This paper presents XingChen-UniFE, a unified large audio language model (LALM) for general speech restoration (GSR) and target speaker extraction (TSE). The proposed model employs complementary semantic and acoustic representations through a dual-stream encoder, where a self-supervised learning model provides hierarchical semantic features and a learnable acoustic branch captures fine-grained structures from complex spectra. A bidirectional cross-attention module further facilitates information exchange between the two representations. The fused representation conditions the LALM for discrete speech token generation with task and speaker information. Experiments demonstrate strong performance on both GSR and TSE and robust generalization to unseen degradation conditions.

Explore related subjects

eess.AScs.SD

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Kelan Kuang, Qihan Hu, Jian Kang, Jie Li, Yongxiang Li, Yi Liu. 2026-10-21. XingChen-UniFE: A Unified Generative Speech Front-End via Complementary Acoustic-Semantic Modeling. https://arxiv.org/abs/2610.29164

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

A Semi-spontaneous Dutch Speech Dataset for Speech Enhancement and Speech Recognition

We present DRES: a 1.5-hour Dutch realistic elicited (semi-spontaneous) speech dataset from 80 speakers recorded in noisy, public indoor environments. DRES was designed as a test set for the evaluation of state-of-the-art (SotA) automatic speech recognition (ASR) and speech enhancement (SE) models in a real-world scenario: a person speaking in a public indoor space with background talkers and noise. The speech was recorded with a four-channel linear microphone array. In this work we evaluate the speech quality of five well-known single-channel SE algorithms and the recognition performance of eight SotA off-the-shelf ASR models before and after applying SE on the speech of DRES. We found that five out of the eight ASR models have WERs lower than 22\% on DRES, despite the challenging conditions. In contrast to recent work, we did not find a positive effect of modern single-channel SE on ASR performance, emphasizing the importance of evaluating in realistic conditions.

eess.AS↗
arXiv2026

Depth through recurrence: Looped transformers for flow-matching TTS

We study how to organize Transformer depth through recurrence in flow-matching text-to-speech, varying the amount, order, and placement of weight reuse. Seven layouts perform 18 block calls per network evaluation under a common training objective and sampler. On Seed-TTS and LibriSpeech-PC, SEQUENCE applies each of nine blocks twice consecutively, retaining competitive intelligibility, speaker similarity, and predicted speech quality at 32 sampling steps with 47.1% fewer parameters than the unshared baseline. Cycling six blocks three times further reduces model size but raises 32-step word error rates relative to cycling nine blocks twice. At matched parameter counts and executed depth, reuse order and sharing position produce different quality trade-offs. These comparisons depend on sampling budget: Prefix and Suffix have similar 32-step word error rates, but Suffix is worse by 3.44 and 5.97 percentage points at four steps on the two datasets, respectively. Only Middle ranks first or second in mean word error rate at 32 and four steps on both datasets. These results show that the organization of recurrent computation affects synthesis quality, and that reuse layouts should be selected for both the sampling budget and the quality dimensions of interest.

eess.AS↗
arXiv2026

An Absolute Acoustic Readout of Plate Wave Speed for Bullion Authentication, Using a Measured Fill-Factor Correction for Struck Coins

Acoustic "ping" testing of bullion is empirical: it compares a coin's ringing frequency with frequencies recorded for the same coin type, so it cannot assess a specimen absent from the reference library. This work replaces the comparison with an absolute measurement, computed from the measured resonance and geometry through a physical model rather than retrieved from a library. Free-plate flexural resonance is calibrated on five machined aluminium discs ($a/t=2.58$--$10.75$), recovering an empirical finite-thickness coefficient $β=0.895\pm0.047$ that reduces the worst-case departure from thin-plate theory from 8.9% to 0.75%. A fill factor, the ratio of a coin's displaced volume to its nominal cylinder, supplies the effective thickness that relief hides from calipers, extending the calibration to struck coins. The readout is the plate wave speed $c_{p}=\sqrt{E/ρ(1-ν^{2})}$, a material constant, conditioned on an assumed Poisson's ratio. On six struck 0.999 copper rounds not used in the calibration it recovers a Young's modulus of $119.1\pm6.7$ GPa, inside the accepted 110 to 128 GPa. The recovered wave speeds form distinct observed groups across the measured specimens, with the strongest independent validation provided by these copper rounds. The $(3,0)/(2,0)$ mode ratio falls below its thin-plate value in thirteen of sixteen specimens (one-sided sign test, $p=0.011$), as finite-thickness theory predicts. Weighing displaced mass to 0.1 mg rather than 0.1 g cut per-specimen volume uncertainty from 1.6% to 30% to 0.5% to 2.6% and shrank an apparent 11% design-dependent bias to a 4.7% residual on one deeply relieved silver design. The readout is now limited by tabulated elastic constants and the assumed Poisson's ratio, not by the acoustics, which are reproducible to 0.15%.

eess.AS↗