Improving biometric voice authentication systems: robustness in facing short-term dysphonies

• Grant number: 21/12407-4
• Support Opportunities: Regular Research Grants
• Start date: March 01, 2022
• End date: February 29, 2024
• Field of knowledge: Engineering - Electrical Engineering
• Principal Investigator: Rodrigo Capobianco Guido
• Grantee: Rodrigo Capobianco Guido
• Host Institution: Instituto de Biociências, Letras e Ciências Exatas (IBILCE). Universidade Estadual Paulista (UNESP). Campus de São José do Rio Preto. São José do Rio Preto , SP, Brazil
• Associated researchers: Fernando Fernandes Paiva ; Ivan Nunes da Silva
• Associated scholarship(s): 22/05186-4 - Improving Biometric Voice Authentication Systems: Robustness in Facing Short-Term Dysphonies, BP.TT

Abstract

Once speech acoustically changes in an involuntary manner, the speaker is stricken by an organic, functional or organic-functional dysphonia. Consequently, their individual acoustic identification, not only by humans but mainly by machines, mighty be at risk. Considering this is an almost unexplored topic, the intention is to investigate how dysphonies affect biometric speaker verification (BSV), creating robust algorithms for that task when the subjects are affected. Particular attention will be dedicated to people with temporary dysphonia, such as hoarseness and cold, which create a barrier to phonation and, consequently, to accurate acoustic analysis. As soon as the literature has been systematically reviewed, the investigative procedure will commence. For feature extraction, the intention is to compare the potential of autoencoder-based feature learning with the analysis provided by handcrafted extraction, such as that obtained with the Discrete-Time Wavelet-Packet Transform (DTWPT), aided by Paraconsistent Feature Engineering (PFE). Then, in order to correctly authenticate the speakers enrolled in the system, the accuracy and performance of Residual Neural Networks (RNNs) and Deep Spiking Neural Networks (DSNNs), among others, will be evaluated and compared taking two modalities into account: text-dependent and text-independent. Lastly, the intention is to publish the results in renowned scientific journals. (AU)

Scientific publications

(4)

(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)

BARBON JUNIOR, SYLVIO; GUIDO, RODRIGO CAPOBIANCO; AGUIAR, GABRIEL JONAS; SANTANA, EVERTON JOSE; PROENCA JUNIOR, MARIO LEMES; PATIL, HEMANT A.. Multiple voice disorders in the same individual: Investigating handcrafted features, multi-label classification algorithms, and base-learners. SPEECH COMMUNICATION, v. 152, p. 14-pg., 2023-07-01. (21/12407-4)

HO, TIN KAM; LUO, YEN-FU; GUIDO, RODRIGO CAPOBIANCO. Explainability of Methods for Critical Information Extraction From Clinical Documents A survey of representative works. IEEE SIGNAL PROCESSING MAGAZINE, v. 39, n. 4, p. 11-pg., 2022-07-01. (21/12407-4)

CONTRERAS, RODRIGO COLNAGO; VIANA, MONIQUE SIMPLICIO; FONSECA, EVERTHON SILVA; DOS SANTOS, FRANCISCO LLEDO; ZANIN, RODRIGO BRUNO; GUIDO, RODRIGO CAPOBIANCO. An Experimental Analysis on Multicepstral Projection Representation Strategies for Dysphonia Detection. SENSORS, v. 23, n. 11, p. 36-pg., 2023-05-30. (21/12407-4, 22/05186-4)

GUIDO, RODRIGO CAPOBIANCO. Wavelets behind the scenes: Practical aspects, insights, and perspectives. PHYSICS REPORTS-REVIEW SECTION OF PHYSICS LETTERS, v. 985, p. 23-pg., 2022-08-27. (21/12407-4)