Implementation of a Voice Recognition System in the Nasa Yuwe Language Based on Convolutional Neural Networks

Julio Enrique Muñoz Burbano

Universidad del Cauca

Pablo Emilio Jojoa Gomez

Universidad del Cauca

Fausto Miguel Castro Caicedo

Universidad Nacional Abierta y a Distancia

Introduction: This paper presents the Implementation of an algorithm for voice recognition in the Nasa Yuwe language based on Convolutional Neural Networks (CNN), developed at the Universidad del Cauca in the year 2022.

Problem: The Nasa Yuwe language is phonetically rich, as it has 32 vowels and 34 consonants, which leads to confusion in pronunciation and therefore difficulties in recognizing voice patterns.

Objective: To implement a speech recognition algorithm for the Nasa Yuwe language supported in CNN.

Methodology: The preprocessing of the audio signals was carried out to subsequently obtain the characteristics through the scalograms of the Mel coefficients. Finally, an architecture of the CNN is proposed for the classification process.

Results: A DataSet is built from the scalograms of the voice patterns, and the CNN training process is carried out.

Conclusion: The implementation of a Voice Recognition System based on CNN provides low margins of error in the word classification process of the Nasa Yuwe language.

Originality: The proposed voice recognition system is the first and only one of its kind that has been carried out so far, with the purpose of collaborating in the process of teaching, preserving and learning the Nasa Yuwe language.

Limitations: It is necessary to increase the number of voice patterns provided by native speakers, and there is a need to implement other technological tools that allow for the conservation and dissemination of the Nasa Yuwe language.

Keywords: VRS (Voice Recognition System), Nasa Yuwe Language, Mel coefficients, machine learning, CNN (Convolutional Neural Network)
Published
2023-01-22
Downloads

How to Cite

[1]
J. E. Muñoz Burbano, P. E. Jojoa Gomez, and F. M. Castro Caicedo, “Implementation of a Voice Recognition System in the Nasa Yuwe Language Based on Convolutional Neural Networks”, ing. Solidar, vol. 19, no. 1, pp. 1–20, Jan. 2023, doi: 10.16925/2357-6014.2023.01.01.
Metrics
File downloads
141
https://plu.mx/plum/a/?doi=10.16925/2357-6014.2023.01.01