Natural speech reconstruction system with bandwidth constraint

Chong, Yee Xiang (2019) Natural speech reconstruction system with bandwidth constraint. Final Year Project, UTAR.

Preview

PDF Download (1958Kb) | Preview

Abstract

Natural speech reconstruction system with bandwidth constraint is a system that enhancing existing video calling system on adaption for an emergency call. This system is mainly focused on the reconstruction of the speech but not focus on the video. The objective of the project is to develop a system that can convert the text to speech, convert back from speech to text based on the speaker's voice and to design a training using smaller dataset while achieving similar accuracy. The system flow starts from a video is being affected by an emergency network, the system will capture the speech from the speaker. The speech will be converted into text and send to the listener device. After the listener device receives the data, the system will convert the text back to speech. This conversion involves the training voice model which is trained early and store inside the voice model database. After the conversion, the speech will be playing by the system to the listener. The technique used in this project is speech recognition, speech synthesis and machine learning. Speech recognition is doing the speech to text conversion in the speaker device. Next, speech synthesis is the technique used to convert text to speech in the listener device. Lastly, machine learning is involved in training a voice model by using supervised machine learning.

Actions (login required)