Pretrained Neural Network models for detecting respiratory conditions through audio analysis

• Grant number: 22/16374-6
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: June 01, 2023
• End date: December 31, 2024
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Marcelo Finger
• Grantee: Marcelo Matheus Gauy
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Company: Universidade de São Paulo (USP). Centro de Inovação da USP (INOVA)
• Associated research grant: 19/07665-4 - Center for Artificial Intelligence, AP.eScience.CPE

Abstract

This project is a continuation of a study (called SPIRA, FAPESP project 2020/06443-5) which aim to create an automated system capable of early detection of respiratory insufficiency through speech analysis.This proposal has as its central objective the development of pretrained neural network models for speech processing of Brazilian Portuguese Speech. These models will be refined on respiratory insufficiency data, as well as perhaps on other comparable classification tasks, with the intention of obtaining efficient models for those tasks despite the availability of only a small quantity of labeled data.Pretrained models have two phases. In the first phase, the model is trained in a large quantity of unlabeled generic audios using generic tasks. In the second phase, these models are refined with labeled data, often only present in small quantities, for a specific task (like the detection of respiratory insufficiency). In this project, we will study and implement the most efficient forms of generic tasks for pretraining, with the goal of extending the results already obtained by the SPIRA project to other possible respiratory conditions and tasks. Since the amount of audios for the refinement phase is small, this will rely on an efficient pretraining scheme to reach good performance on multiple tasks.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

BERTI, LARISSA CRISTINA; GAUY, MARCELO; DA SILVA, LUANA CRISTINA SANTOS; RIOS, JULIA VASQUEZ VALENCI; MORAIS, VIVIAM BATISTA; ALMEIDA, TATIANE CRISTINA DE; SOSSOLETE, LEISI SILVA; QUIRINO, JOSE HENRIQUE DE MOURA; MARTINS, CAROLINA FERNANDA PENTEAN; FERNANDES-SVARTMAN, FLAVIANE R.; et al. Acoustic Characteristics of Voice and Speech in Post-COVID-19. HEALTHCARE, v. 13, n. 1, p. 14-pg., 2025-01-01. (23/00488-5, 22/16374-6)