An HCI Speech-Based Architecture for Man-To-Machine and Machine-To-Man Communication in Yorùbá Language

Abstract

Man communicates with man by natural language, sign language, and/or gesture but communicates with machine via electromechanical devices such as mouse, and keyboard.  These media of effecting Man-To-Machine (M2M) communication are electromechanical in nature. Recent research works, however, have been able to achieve some high level of success in M2M using natural language, sign language, and/or gesture under constrained conditions. However, machine communication with man, in reverse direction, using natural language is still at its infancy. Machine communicates with man usually in textual form. In order to achieve acceptable quality of end-to-end M2M communication, there is need for robust architecture to develop a novel speech-to-text and text-to-speech system.

In this paper, an HCI speech-based architecture for Man-To-Machine and Machine-To-Man communication in Yorùbá language is proposed to carry Yorùbá people along in the advancement taking place in the world of Information Technology. Dynamic Time Warp is specified in the model to measure the similarity between the voice utterances in the sound library. In addition, Vector Quantization, Guassian Mixture Model and Hidden Markov Model are incorporated in the proposed architecture for compression and observation. This approach will yield a robust Speech-To-Text and Text-To-Speech system.

Keywords: Yorùbá Language, Speech Recognition, Text-To-Speech, Man-To-Machine, Machine-To-Man