analiza aiqbench
This study presents the best-performing raw waveform acoustic models on the TIMIT dataset, achieving a phone error rate of 13.9% on the development set by combining SincNet or CNN architectures with Bidirectional LSTMs. The analysis reveals that transfer learning from WSJ significantly boosts consonant recognition while highlighting that dominant confusion patterns stem from inherent phonetic similarities rather than architectural differences. For practitioners, the key takeaway is that BLSTM layers are particularly effective for transition-dependent phonetic classes, offering a clear path for optimizing raw waveform systems without relying on filterbank features.
Phonetic Error Analysis of Raw Waveform Acoustic Models
arXiv:2606.07030v1 Announce Type: cross Abstract: We analyse error patterns of raw waveform acoustic models on TIMIT phone recognition beyond the overall phone error rate (PER). PER is decomposed across three broad phonetic class (BPC) categorisations, and confusion matrices are constructed from substitution errors. Our models combine parametric (SincNet, Sinc2Net) or non-parametric CNNs with Bidirectional LSTMs, achieving 13.9%/15.3% PER on Dev/Test, the best reported results for raw waveform models on TIMIT. Transfer learning from WSJ reduces PER to 11.3%/12.3%, surpassing the Filterbank baseline. Per-BPC analysis reveals that BLSTM layers benefit transition-dependent classes most, while WSJ transfer learning improves consonants roughly three times more than vowels. Confusion patterns are consistent across raw waveform and Filterbank systems, indicating that the dominant confusions reflect inherent phonetic similarities.
Abstract:We analyse error patterns of raw waveform acoustic models on TIMIT phone recognition beyond the overall phone error rate (PER). PER is decomposed across three broad phonetic class (BPC) categorisations, and confusion matrices are constructed from substitution errors. Our models combine parametric (SincNet, Sinc2Net) or non-parametric CNNs with Bidirectional LSTMs, achieving 13.9%/15.3% PER on Dev/Test, the best reported results for raw waveform models on TIMIT. Transfer learning from WSJ reduces PER to 11.3%/12.3%, surpassing the Filterbank baseline. Per-BPC analysis reveals that BLSTM layers benefit transition-dependent classes most, while WSJ transfer learning improves consonants roughly three times more than vowels. Confusion patterns are consistent across raw waveform and Filterbank systems, indicating that the dominant confusions reflect inherent phonetic similarities.
Sound (cs.SD); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
arXiv:2606.07030 [cs.SD]
(or arXiv:2606.07030v1 [cs.SD] for this version)
https://doi.org/10.48550/arXiv.2606.07030 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 221 cuvinte
Subiecte conexe
A Generative Grammar Underlying the Voynich Manuscript, the Pastiche Hypothesis: Evidence from Large Language Models
arXiv · 21 sept. 2026
FuseAlign: Forced Alignment in the Wild
arXiv · 30 sept. 2026
Looking for Affect in Spontaneous Finnish Speech through Linguistic Interpretability
arXiv · 28 iul. 2026
PunGraph: Retrieval-Enhanced Phonetic-Semantic Graph Reasoning for Pun Understanding
arXiv · 16 sept. 2026