Afan Oromo Speech-Based Computer Command And Control: An Evaluation With Selected Commands

Abstract

Speech-based computer command and control uses natural speech to make a computer listen to what humans say in their language and perform tasks like commands. However, there is no any Speech-based command and control system studied and developed for Microsoft Word in Afan Oromo. The main objective of this study is to investigate and develop Afan Oromo Speech-Based Command and Control system using some selected Command and Control words from MS Word. In order to achieve this objective; a Small vocabulary, isolated word, Speaker independent, and HMM-based speech recognizer was developed using the HTK toolkit. To develop this Automatic speech-based computer command first the translation of selected MS command words from English to Afan Oromo command words is done. The audio was recorded from 38 speakers (16 female and 22 male) aged between 18 and 40, based on their availability. Word-level speech recognition is performed using MFCC and data processing, which are the most generic and effective approaches to speech recognition. Out of a total of 64 MS, command words, 54 words (84.37%) were used for training and 10 words (15.63%) were used for testing. Live and non-live evaluation techniques were used to evaluate the recognizer's performance. Since neighbors' phones affect non-live recognizer, live with variable environment speech recognizer performs better than non-live recognizer. The performance of the mono-phone tied state tri-phone and tri-phone-based recognizer was 78.12%, 86.87%, and 88.99% respectively. Thus, the tri-phone-based recognizer has the best performance than others in non-live recognizers. The challenges of resource limitations in this research study are restricted to investigating speech-based commands for computers with only selected MS commands, which are very significant keys for text processing. To evaluate a speech-based interface in the real environment there are no components of object-as-a-service. The experimental findings of this study demonstrated that, if the required size of language resources is available the computer based Afan Oromo speech-based interface for the command-and-control purpose could be developed.