Differences

This shows you the differences between two versions of the page.

Link to this comparison view

unit_selection_speech_synthesizer_for_the_polish_language [2014/03/18 21:01]
bziolko created
unit_selection_speech_synthesizer_for_the_polish_language [2014/03/18 21:02] (current)
bziolko
Line 1: Line 1:
 +Sebastian Dziadzio (Poland)
  
 +**Unit selection speech synthesizer for the Polish language**
 +
 +Student paper
  
 Over the last 80 years, due to advancements in computer science and linguistics, synthesised speech went a long way from low quality, unintelligible and audibly artificial to easily comprehensible and almost natural-sounding. The paper presents a unit selection text-to-speech system based on the Festival framework created at the University of Edinburgh Centre for Speech Technology Research. The introductory part consists of a concise overview of synthesis history, recent techniques, and existing commercial systems. Thereafter, relevant mathematical and linguistic terms are defined and briefly discussed. The main part is a detailed technical report, illustrating consecutive stages of the development process: corpus design and recording, labeling, building utterance structures for the database, pre-calculation of acoustical distances between units, preparing cluster trees and building voice description and documentation. In the conclusion, performance tests results are presented, along with a discussion and development plans. Over the last 80 years, due to advancements in computer science and linguistics, synthesised speech went a long way from low quality, unintelligible and audibly artificial to easily comprehensible and almost natural-sounding. The paper presents a unit selection text-to-speech system based on the Festival framework created at the University of Edinburgh Centre for Speech Technology Research. The introductory part consists of a concise overview of synthesis history, recent techniques, and existing commercial systems. Thereafter, relevant mathematical and linguistic terms are defined and briefly discussed. The main part is a detailed technical report, illustrating consecutive stages of the development process: corpus design and recording, labeling, building utterance structures for the database, pre-calculation of acoustical distances between units, preparing cluster trees and building voice description and documentation. In the conclusion, performance tests results are presented, along with a discussion and development plans.
Copyright © XXII PVC Organizing Committee 2013. All Rights Reserved.