Details of Grant 

EPSRC Reference:	GR/S21281/01
Title:	Switching Linear Dynamical Models for Automatic Speech Recognition
Principal Investigator:	King, Professor S
Other Investigators:
Researcher Co-Investigators:
Project Partners:
Department:	Centre for Speech Technology Research
Organisation:	University of Edinburgh
Scheme:	Standard Research (Pre-FEC)
Starts:	01 July 2003	Ends:	30 June 2006	Value (£):	193,502
EPSRC Research Topic Classifications:	Human Communication in ICT
EPSRC Industrial Sector Classifications:	Communications
Related Grants:
Panel History:
Summary on Grant Application Form
This proposal concerns novel acoustic models for automatic speech recognition (ASR) -- that is, alternatives to Hidden Markov Models (HMMs). Performance of ASR systems based on Hidden Markov Models (HMMs) has reached a plateau in recent years, with only small, incremental improvements being made. In order to build ASR systems with greater accuracy than the current state of the art, we believe alternatives to HMMs must be investigated. One possibility that we have been working on recently is the Linear Dynamical Model, and the research proposed here would significantly enhance the modelling power of this model, leading to higher accuracy speech recognition. The Linear Dynamical Model (LDM) is a continuous state model, in contrast to the HMM which has a discrete hidden state. This reflects our knowledge of speech production - the trajectories of the articulators are continuous and smooth, not discrete. However, the acoustic speech signal exhibits a number of non-continuous features (e.g. stops and vowel-fricative transitions) which the LDM must deal with. Additionally, speech is multi-modal (e.g. some speech sounds vary with context). Our proposal addresses these issues by introducing a switching process which controls the parameters of the LDM. The main problem we will address is finding ways of determining the nature of this process.
Key Findings
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Potential use in non-academic contexts
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Impacts
Description This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk Summary Date Materialised
Sectors submitted by the Researcher
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Project URL:
Further Information:
Organisation Website:	http://www.ed.ac.uk