Balvin

                                                                        Abstract:

Reliable phoneme classification is fundamental to intelligent hearing-aid systems because it enables adaptive amplification, environmental noise suppression, and speech enhancement, thereby improving speech intelligibility for individuals with hearing impairment. Previous studies have shown that Mel Frequency Cepstral Coefficients (MFCCs) combined with non-parametric k-Nearest Neighbors (k-NN) classifiers can achieve competitive speech classification performance with relatively low computational complexity. However, recognition accuracy decreases for short-duration phonemes, and limited training data can restrict model generalization. To overcome these limitations, a hybrid speech classification architecture was developed by combining MFCC-based acoustic feature extraction, deep feature representation using a Convolutional Neural Network (CNN), and weighted k-Nearest Neighbour (k-NN) decision making. The proposed methodology was implemented within the MATLAB environment and experimentally validated using the TIMIT Acoustic–Phonetic Continuous Speech Corpus. System performance was quantified through recall, classification accuracy, F1 score, precision, confusion matrix evaluation, Receiver Operating Characteristic (ROC) analysis, and computational efficiency. The experimental evaluation yielded an overall phoneme recognition accuracy of 96.4%, while achieving 82.8% accuracy for short-duration phonemes, demonstrating the effectiveness of the proposed hybrid learning strategy. It outperformed conventional HMM, SVM, ANN, standalone CNN, and MFCC-based k-NN classifiers while maintaining computational efficiency suitable for embedded implementation. The experimental findings indicate that the proposed hybrid framework substantially improves phoneme discrimination and establishes a scalable, computationally efficient architecture capable of supporting future real-time intelligent hearing-aid applications.

Biography:

Dr. Balvin Owen Thorpe, Ph.D., PE, MSc, CEM, SMIEEE, MJIE .

Associate Professor Balvin Owen Thorpe is a distinguished engineering leader, academic, and researcher with over three decades of experience spanning telecommunications, energy systems, and higher education. He currently serves as the Vice Dean of the Faculty of Engineering and Computing at the University of Technology, Jamaica, where he provides leadership in quality assurance, accreditation, and research development. Dr. Thorpe holds a Ph.D. in Electrical Engineering from the University of Technology, Jamaica, as well as an MSc degree in Curriculum and Instruction from Florida International University, and an MSc in Digital Technology from the University of the West Indies, Mona. He is also a Certified Energy Manager and an ISO 9001:2015 Quality Management Systems Certified professional. Along with his academic leadership, Dr. Thorpe has a notable industry background, having served as Regional Network Development Engineering Manager at Cable & Wireless Jamaica, where he managed multi-million-dollar telecommunications infrastructure projects throughout the Caribbean. A Senior IEEE Member and Professional Engineer (Jamaica), Dr. Thorpe is deeply committed to professional service. He served as the student counselor for the UTech student chapter for 10 years. He also served as the Project lead for the IEEE Jamaica Section for 3 years. Dr. Thorpe is the current President of the Jamaica Institution of Engineers (JIE. His research interests include wireless communications, smart grids, renewable energy, and AI-based speech processing, with publications in IEEE and regional journals. Dr. Thorpe is recognized for his ability to bridge academia, industry, and professional engineering bodies to drive innovation and sustainable development.

Email:

Professor of IEEE Jamaica Section

Presenter:

Dr. Balvin O. Thorpe,

Senior Member, 

IEEE Vice Dean,

 Faculty of Engineering & Computing University of Technology, Jamaica

Biography:

Presenter: Dr. Balvin O. Thorpe Senior Member, IEEE Vice Dean, Faculty of Engineering & Computing University of Technology, Jamaica

Email: