We describe and analyze a simple and effective algorithm for sequence segmentation applied to speech processing tasks. We propose a neural architecture that is composed of two modules trained jointly: a recurrent neural network (RNN) module and a structured prediction model. The RNN outputs are considered as feature functions to the structured model. The overall model is trained with a structured loss function which can be designed to the given segmentation task. We demonstrate the effectiveness of our method by applying it to two simple tasks commonly used in phonetic studies: word segmentation and voice onset time segmentation. Results suggest the proposed model is superior to previous methods, obtaining state-of-the-art results on the tested datasets.
Download full-text PDF |
Source |
|---|---|
| http://www.ncbi.nlm.nih.gov/pmc/articles/PMC5638122 | PMC |
| http://dx.doi.org/10.1109/ICASSP.2017.7952591 | DOI Listing |
Publication Analysis
Top Keywords
Similar Publications
Neuronal Imaging at 8-Bit Depth to Combine High Spatial and High Temporal Resolution With Acquisition Rates Up To 40 kHz.
J Biophotonics
January 2025
Univ. Grenoble Alpes, CNRS, LIPhy, Grenoble, France.
A challenge in neuroimaging is acquiring frame sequences at high temporal resolution from the largest possible number of pixels. Measuring 1%-10% fluorescence changes normally requires 12-bit or higher bit depth, constraining the frame size allowing imaging in the kHz range. We resolved Ca or membrane potential signals from cell populations or single neurons in brain slices by acquiring fluorescence at 8-bit depth and by binning pixels offline, achieving unprecedented frame sizes at kHz rates.
View Article and Find Full Text PDFPharyngeal airway dimensions and adipose distribution in the minipig.
J Oral Biol Craniofac Res
December 2024
Department of Orthodontics, University of Washington, Seattle, USA.
Objective: To evaluate the pharyngeal airway dimensions and regional pharyngeal adipose distribution in the young adult minipig model.
Materials And Methods: Eight 7-8-months-old Yucatan minipigs, half male and female, were sedated and placed prone to scan the pharyngeal region. Magnetic resonance imaging (MRI) was performed using dynamic turbo-field echo (TFE)-sequence with respiratory gating and adipose-weighted sequence.
Identification of f.a., sp. nov. and large indels in the rRNA cistron that split the genus.
Int J Syst Evol Microbiol
January 2025
National Fungal Culture Collection of India (NFCCI), Biodiversity and Palaeobiology Group, MACS-Agharkar Research Institute, Pune, India.
Yeast strains representing a novel asexual ascomycetous species were isolated from seven flowers. Sequencing of the chromosomal regions coding for the D1/D2 domains of the large subunit ribosomal RNA, the ITS1-5.8S-ITS2 segments and parts of the gene coding for the small subunit ribosomal RNA showed that the isolates were conspecific.
View Article and Find Full Text PDFBackground: Alzheimer's disease is a common neurodegenerative disease that affects the lives of millions of people worldwide. In our study, we aim to evaluate the effectiveness and accuracy of the latest automated brain volume analysis method for GM and WM analysis in health control and patients with AD and to compare optimized cut-off values for both regions between subjects.
Method: The original baseline scans from 37 HC (HC) and 39 mild AD patients downloaded from ADNI.
Background: Renal atrophy may reflect an end organ consequence of chronic vascular disease. Renal volume loss may therefore provide a window into brain aging and Alzheimer disease risk.
Method: We obtained whole-body 1.
Want AI Summaries of new PubMed Abstracts delivered to your In-box?
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!