COMP47700 Speech and Audio

Academic Year 2023/2024

This module provides an introduction to speech and hearing and how digital representations of speech and audio can be processed. The area of speech and audio processing has evolved rapidly with applications like Voice over IP (e.g. Skype), media streaming (e.g. Spotify, YouTube) using encoding to optimise network bandwidth resources while maintaining user's the quality of experience. Advances in automatic speech recognition and speech synthesis have allowed speech controlled applications (e.g. Siri, Alexa) possible.

Data understanding is an important step in preprocessing and feature selection for predictive modelling. Machine learning models for applications such as automatic speech recognition, speech analysis and quality of experience prediction rely on understanding the salient properties of discrete time series signals.

This module provides theoretical grounding, practical knowledge, and hands on experience using high level programming languages (e.g. Python) to allow students to apply domain knowledge from speech and audio to data driven analytics. Fundamental analysis and processing methods for speech and audio applications will be explored. This module takes a hands-on approach making use of third-party libraries and APIs.

The module covers the following core topics:
• Introduction to speech and audio processing
• Basic audio processing
• Speech
• The human auditory system
• Psychoacoustics
• Speech communications
• Audio analysis
• It also introduces advanced topics and current research within the field of speech and audio processing. For example:
– Psychoacoustic modelling
– Speech and Audio Quality
– Objective models of intelligibility and quality
– Speaker recognition
– Spatial audio