Please use this identifier to cite or link to this item: http://hdl.handle.net/10603/222031
Title: Significance of incorporating excitation source features for Improved speech emotion recognition
Researcher: Pravena.D
Guide(s): Govind.D , Soman K.P
Keywords: Engineering and Technology,Computer Science,Computer Science Interdisciplinary Applications
Speech recognition; Speech emotion recognition; Emotion analysis
University: Amrita Vishwa Vidyapeetham (University)
Completed Date: 08/2018
Abstract: The objective of the work presented in the thesis work is to demonstrate the effectiveness of the excitation parameters for improving the recognition performances of the simulated emotions. The strength of excitation (SoE) and instantaneous F0 parameters are selected as the salient excitation source parameters for the present work. Due to unavailability of large emotion databases in terms of number speakers, number of sessions and number of languages, a fairly large database with multiple speakers in three Indian languages ( Malayalam, Tamil and Indian English) for the distinct emotions such as Anger, Happy and Sad (apart from Neutral) is developed. During the development of the database, the emotions elicited using emotionally biased prompts and emotionally neutral prompts are observed to show varied emotion recognition rates by the same speakers. The emotions elicited using the emotionally biased prompts showed better emotion discrimination as compared to emotionally neutral prompts. Therefore, the database developed with emotionally biased prompts in all the three languages is chosen for the experiments presented in this thesis. For the excitation source analysis of various emotions, each emotive utterance in the database is recorded with simultaneous speech and electroglottogram (EGG) signals. The features estimated from EGG signals are used as the ground truth for the comparisons with corresponding features estimated from speech signals. Since SoE and instantaneous F0 parameters are the emotion dependent parameters used in the present work, the accurate estimation of these parameters are essential for the emotive speech utterances...
Pagination: XXV, 148
URI: http://hdl.handle.net/10603/222031
Appears in Departments:Center for Computational Engineering and Networking (CEN)

Files in This Item:
File Description SizeFormat 
01_title.pdfAttached File142.3 kBAdobe PDFView/Open
02_certificate.pdf142.8 kBAdobe PDFView/Open
03_declaration.pdf72 kBAdobe PDFView/Open
04_dedicated.pdf43.31 kBAdobe PDFView/Open
05_contents.pdf89.84 kBAdobe PDFView/Open
06_acknowledgements.pdf95.44 kBAdobe PDFView/Open
07_list of figures.pdf148.35 kBAdobe PDFView/Open
08_list of tables.pdf120.94 kBAdobe PDFView/Open
09_list of acronyms.pdf72.4 kBAdobe PDFView/Open
10_list of symbols.pdf142.1 kBAdobe PDFView/Open
11_abstract.pdf94.63 kBAdobe PDFView/Open
12_chapter 1.pdf103.05 kBAdobe PDFView/Open
13_chapter 2.pdf438.88 kBAdobe PDFView/Open
14_chapter 3.pdf570.72 kBAdobe PDFView/Open
15_chapter 4.pdf444.45 kBAdobe PDFView/Open
16_chapter 5.pdf256.99 kBAdobe PDFView/Open
17_chapter 6.pdf334.94 kBAdobe PDFView/Open
18_references.pdf122.89 kBAdobe PDFView/Open
19_publications.pdf82.35 kBAdobe PDFView/Open


Items in Shodhganga are protected by copyright, with all rights reserved, unless otherwise indicated.

Altmetric Badge: