Speech Recognition Algorithm of Major Nigerian Languages (Yoruba, Hausa, Ibo) Using K-NN
No Thumbnail Available
Date
2023-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Lead City University
Abstract
This study addresses the crucial need for enhanced voice recognition systems in the realm of human-machine interfaces, particularly with a focus on accent identification algorithms, and their application in the context of Nigerian English speakers. The research aims to improve the accuracy and efficiency of speech recognition for these major Nigerian languages using KNN to increase the efficiency and accuracy of accent identification by using a trained data with the three major Nigerian language. Voiced audio samples of speakers from these tribes speaking English and various platforms such as news media and radio recordings was scrapped and recorded and extracted. The audio data was then preprocessed and transformed from the time domain to the frequency domain using the Fourier transform. Matlab R2015A was employed for model training, encompassing input reading, window size and hop size definition, and noise reduction techniques such as high-pass filtering and spectral subtraction. For feature extraction, Mel Frequency
Cepstral Coefficients (MFCC) were computed for each audio frame, subsequently aggregated to create fixed-length representations for each dialect sample for about sixty seconds in order to ensure uniformity in the inputs. The model underwent training with a classification algorithm KNN, followed by evaluation, which gave an accuracy rate of 84%. This result indicates that the model proficiently predicts the dialects within the context of English speech. The study's outcomes signify substantial progress in the development of an accent detection model tailored to the major Nigerian tribes: Yoruba, Hausa, and Igbo. The research is a significant stride toward more precise and effective voice recognition systems for Nigerian English speakers, contributing to the broader advancement of human-machine interfaces in an increasingly technology-driven world. It is recommended that future
research explores alternative feature extraction techniques, particularly deep learning-based approaches capable of automatically
learning relevant features from raw audio data.
Keywords: Accent. Accuracy, Algorithm, Dialect detection, Feature extraction, Fourier transform, Performance, Speech recognition,
Voice recognition
Word Count: 298 Words
Description
Keywords
Accent. Accuracy, Algorithm, Dialect detection, Feature extraction, Fourier transform, Performance, Speech recognition