Postgraduate Course: Speech Synthesis (LASC11062)
|School||School of Philosophy, Psychology and Language Sciences
||College||College of Arts, Humanities and Social Sciences
|Credit level (Normal year taken)||SCQF Level 11 (Postgraduate)
||Availability||Available to all students
|Summary||This course explores issues in text-to-speech synthesis by taking a detailed look at the theory and practice of state of the art speech synthesis systems. Through lectures students will learn the theory of speech synthesis. In the lab sessions and coursework students will learn about the practical application of this theory as they design, build, and evaluate their own synthetic voice. The syllabus starts from unit selection approaches then builds up to the current state of the art using neural networks. Other topics covered include: creating the data required for unit selection or for training a neural network, speech signal processing, and evaluating speech synthesis.
The course is delivered as a combination of lectures, flipped classrooms, an online forum, short videos, readings, and a practical exercise in the lab.
In the lab, students build their own fully-functional speech synthesis voice, within the Festival framework.
Syllabus: approaches to speech synthesis, text selection and recording data for corpus based approaches, searching inventories for unit selection approaches, prosody, pitch tracking and pitch marking, speech coding and vocoding for speech synthesis, statistical parametric speech synthesis using Hidden Markov models, statistical parametric speech synthesis using Deep Neural Networks, evaluating speech synthesis.
Information for Visiting Students
|High Demand Course?
Course Delivery Information
|Academic year 2022/23, Available to all students (SV1)
|Learning and Teaching activities (Further Info)
Lecture Hours 27,
Programme Level Learning and Teaching Hours 2,
Directed Learning and Independent Learning Hours
|Assessment (Further Info)
|Additional Information (Assessment)
||Comments will be provided on submitted coursework and a structured marking scheme will be used. All students will have the opportunity for an individual 15 minute summative feedback session with the lecturer covering the coursework assessment, after it is returned.
Class-wide formative feedback will be provided based on work previously submitted for Speech Processing (from multiple students, anonymised),
since the style and format is very similar to that required for this course. The event will take the form of a video lecture and blog post, made available by Block 3 Week 5, with follow-up questions answered via the course forum and in-person in the lab.
|No Exam Information
On completion of this course, the student will be able to:
- understand the speech synthesis process, and be familiar with the processing steps required to convert text to speech and be familiar with the different speech synthesis methods currently used by speech synthesis systems and understand the advantages and disadvantages of each
- have a detailed understanding of the principles of unit selection speech synthesis, and the issues involved with choosing suitable candidate units to match a given target sequence and understand the design issues associated with recording data suitable for building a unit selection voice
- have the practical experience of having built a synthetic voice themselves
- be familiar with the different speech coding techniques that can be used for speech synthesis, and understand how these can be used to aid the joining of individual speech segments and how using different signal processing techniques to manipulate speech synthesis output affects the speech quality
- be in a position to discuss current issues in speech synthesis and see where speech synthesis research is heading in the future
|Graduate Attributes and Skills
||Ability to use the industry-standard speech synthesis toolkit, Festival
Ability to make high-quality recordings of speech
Ability to build and tune a unit selection voice
Experimental design and analysis
|Keywords||text-to-speech synthesis,speech signal processing,statistical modelling using Hidden Markov models
|Course organiser||Prof Simon King
Tel: (0131 6)51 1725
|Course secretary||Ms Sasha Wood