Generating Expressive Degree of Emotion in Neutral Speech
Manjare Chandraprabha A1, Shirbahadurkar Suresh D2, Patil Prerna R3
1Manjare Chandraprabha Anil, M.E.(Electronics), Marathwada University, Shivaji University, Aurangabad (Maharashtra), India.
2Dr. Shirhahadurkar, Ph.D.(EC), Marathwada University, Dr. B.A.M. University, Aurangabad, Maharashtra, India.
3Prerna R. Patil, M.E.(Digital Systems) degree from Savitribai Phule Pune University, (Maharashtra), India.
Manuscript received on 15 August 2015 | Revised Manuscript received on 25 August 2015 | Manuscript Published on 30 August 2015 | PP: 33-37 | Volume-4 Issue-6, August 2015 | Retrieval Number: F4153084615/15©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: This paper proposes a statistical phrase/accent model for speech synthesis. In recent years, the work on expressive speech has increased rather than basic emotions. Our aim is to obtain expressive speech from neutral speech. In this proposed method there are two components one is phrase and other is accent. Expectation-Maximization algorithm is used to train statistical speech data. The output generated by proposed method is compared with TD-PSOLA method. The results generated from proposed work is better than TD-PSOLA method.
Keywords: Intonation Modeling, Accent/Phrase, Statistical Parametric Speech Synthesis, TD-PSOLA.
Scope of the Article: Machine Learning