Please use this identifier to cite or link to this item:
|Title:||Voice morphing||Authors:||Sowmya Uthirapathi||Keywords:||DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing||Issue Date:||2009||Abstract:||Voice morphing is a technique to modify source speaker's speech utterance to sound as if the target speaker has spoken it. There are many voice morphing methods but they all have the basic three components 1. Analysis 2. Transformation 3. Conversion Speech coding is used to analyze the speech signal. The goal of speech coding is to represent the speech in digital form with as little bits as possible. And there is a tradeoff between bit rate and voice quality. Aim of this study is to identify an analysis technique, which can generate the synthetic speech with better quality to meet the objectives of speech coding techniques. In this study CELP coding techniques is identified as the best for the analysis of original speech signal when compared to the LPC 10 decoder. The synthetic speech generated by CELP coding technique has better quality and it is closer to the original speech, whereas the synthetic speech generated from the LPC 10 decoder has less quality, more noise, and also unnatural. This is because of the jitter caused by the voiced excitation. So we can choose CELP coding technique for the analysis of speech signal which is the first stage in Voice Morphing. The prediction parameters obtained from CELP analysis were converted into Line Spectral Frequencies. These LSF are used to represent the spectral envelope of source and target speech.||Description:||71 p.||URI:||http://hdl.handle.net/10356/46802||Rights:||Nanyang Technological University||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Theses|
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.