Papers for

voice recognition engineers

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Speech representations lose regional accents between read and spontaneous talking

Do Speech Representations Preserve Regional Accent Across Read and Spontaneous Speech?

Abstract: Regional accent cues can be captured under matched conditions, but it remains unclear whether they persist between read and spontaneous speech. We study RVG1, with 500 German speakers from nine regions, comparing ten speech representations on regional classification and continuous geolocation under matched conditions and speaker-independent read--spontaneous transfer. Whisper performs best under matched conditions, reaching 0.489 nine-way UAR and 148 km median geolocation error, but drops to 0.11/0.18 UAR across transfer directions and 363 km geolocation error. Self-supervised models show a similar degradation, whereas speaker embeddings are less discriminative in-domain but more robust under transfer. This contrast is consistent across classification and geolocation. Across representations, robustness is associated with how little a representation shifts between styles (style-invariance), for which crossstyle speaker retrieval is an interpretable proxy. Age, sex, sentence-overlap, and duration controls do not account for the gap, although channel characteristics contribute. These results show that strong matched-condition performance does not indicate robust regional information.

Mon 5 OctComputation and LanguageSound
The gist
It can be hard to tell where someone is from just by how they speak, especially when they switch between reading text aloud and talking naturally. The authors tested different ways to analyze speech to see if regional accents stay recognizable across these speaking styles. They found that while some methods work well when the speaking style is the same, they struggle when matching read speech to spontaneous speech. The best methods for spotting regions in one style didn’t hold up when the style changed, showing that regional accents are not always preserved in speech analysis.
Open → 2610.06430v1

Geometry conditioned encoders improve long sequence timing retrieval accuracy

Geometry-Conditioned Fixed-Scaffold Encoders for Time-Warp Robust Sequence Retrieval

Abstract: Embedding-based retrieval is attractive for long sequence collections because each item can be encoded once and searched by nearest-neighbor ranking. The difficulty is that the objects being indexed are often observed under a noncanonical clock: cardiac cycles stretch with rate, speech changes with tempo, and sensor traces reach comparable states at different speeds. This paper studies a specific source of instability in patch-based encoders for this regime. If patch boundaries are chosen from signal geometry, then the tokenization can change under the same temporal deformation that the representation is expected to tolerate. We propose GeoPatch, a fixed-scaffold patch encoder that keeps token support independent of geometry and uses slope, curvature, acceleration, affine-residual, and confidence descriptors only as continuous conditioning variables. The design turns boundary variation into feature modulation: geometry can change the embedding through a controlled pathway, but it cannot change the number, order, or support of local tokens. We formalize this distinction through a mechanism-level stability analysis that separates boundary drift, affine timing variation, confidence-weighted geometry perturbation, and retrieval-margin effects. The same local tokens support global embedding retrieval and late-interaction scoring, so the scoring rule can be matched to the evaluation protocol. Across ECG, speech, and multivariate time-series retrieval tasks, GeoPatch improves early-rank retrieval under timing variation while exposing a clear trade-off between local surface matching and strict non-overlap retrieval.

Tue 29 SeptArtificial Intelligence
The gist
Matching long signals like heartbeats or speech recordings is hard because they can speed up or slow down, changing how they look over time. The authors studied how breaking these signals into pieces for analysis can accidentally change when the signal is warped by speed changes. They designed a new method called GeoPatch that keeps the pieces fixed and only adjusts the details, making it better at recognizing signals even if their timing changes. This approach was tested on heart, speech, and other time-series data with promising improvements in finding matches quickly.
Open → 2609.36809v1