MIMIC HUMAN SPEECH IN BAHASA INDONESIA USING SPEECH RECOGNITION AND SPEECH SYNTHESIS

Abstract

People use speech recognition and speech synthesis to help, support, and boost their daily activities. With just one of the speech technologies, developer can produce various software. Combine both speech technologies, developer could produce more various software. One of the combinations is mimic human speech. This research will discuss about Speech Recognition that use Convolutional Neural Network as machine learning model and Speech Synthesis that use Concatenative Synthesis with syllables as speech unit. Different with the recent related works, this research has simpler approach to mimic speech in Bahasa Indonesia. The purpose of this research is to develop applications to collect, train, and mimic speech in Bahasa Indonesia. User can participate to record their speech. Those speeches are collected to be trained for recognizing speech in the application later. With the trained model, now user is able to make the computer mimic their speech. First, user must identify their speech to be recognized by the application. This step is necessary to create the user digital speech. After that, based on registered syllables, which the speech has been identified by the application, user is able to generate speech by making sentences from those syllables. The applications to collect and mimic speech are developed as web-based application and the application to train is developed as text-based application.