Skip to main navigation Skip to search Skip to main content

A Convolutional Speaker Recognition System for Modulated NBFM

  • José-María Jimenez-Coronado
  • , Emanuel Hernández-Cepeda
  • , Roger-de-Jesús Morales-Monge
  • , Felipe Meza-Obando

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper presents a novel approach for speaker recognition in narrowband FM (NBFM) radio communication systems, exploiting raw in-phase and quadrature (IQ) signals from the receiver instead of using demodulated audio. The solution to the user validation problem in critical public safety systems is solved with a convolutional neural network architecture that classifies speakers based on spectro-temporal features extracted from the IQ components from the modulated signal. The results are encouraging, with the solution achieving an accuracy of 96.20 percent for the spectrogram-based model and 96.75 percent for the Mel frequency cepstral coefficients (MFCC) based model in high signal-to-noise (SNR) scenarios.
Translated title of the contributionUn sistema de reconocimiento de altavoces convolucional para NBFM modulado
Original languageEnglish
Title of host publication2025 IEEE 7th International Conference on BioInspired Processing (BIP)
PublisherIEEE
Pages1-6
Number of pages6
DOIs
StatePublished - 3 Dec 2025

Fingerprint

Dive into the research topics of 'A Convolutional Speaker Recognition System for Modulated NBFM'. Together they form a unique fingerprint.

Cite this