Speech Clarity Analysis for ALS Patients Using CNN-BiGRU and Speaker-Relative Regression

Read full text Summarise with VIRAT
Authors: Payal Katiyar, Rishika Raj, Thilagavathy R
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper proposes a regression-based framework for estimating speech clarity in patients with Amyotrophic Lateral Sclerosis (ALS) utilizing a hybrid CNN-BiGRU architecture. In contrast to traditional binary classification methods, speech clarity is represented as a continuous variable to more accurately reflect the progressive characteristics of dysarthria. Experiments utilizing the TORGO dysarthric speech dataset illustrate that conventional classification techniques may produce deceptive performance in realistic scenarios. To deal with differences between speakers, a speaker-relative normalization scheme is introduced. This makes it easier to keep track of how speech quality changes over time. The suggested method leads to better correlation-based evaluation metrics, which shows that it could be useful for keeping an eye on the progression of ALS in a clinical setting.

Fields of Study & Subjects

Voice and Speech Disorders Speech Recognition and Synthesis Speech and Audio Processing

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
25 March 2026
Language
EN
Document type
Preprint