Vitae, Andrew Cameron Morris, October
Associate at IRCAM, on project VIVOS,
working on expressive speech synthesis for use in T2S for computer
animation and voice adaptation for possible use in the entertainment
industry (games and
- 1992 PhD: Doctorate of the Institute
Polytechnique de Grenoble, "An information-theoretical study of speech
processing in the peripheral auditory system and cochlear nucleus:
to the recognition of French voiced stop consonants" (grade "Tres
- 1988 B.A. Hons. in mathematics (grade:
- 1979 RIBA Part 1 (BA) architecture
- 1973 A level Maths (1), further maths
maths (1), physics (2)
- 1973 Driving licence, car and motorbike
years), C++(3 years), Java
(6 months), Pascal, Fortran, Linux and Windows operation systems.
- Expressive speech synthesis
- Natural language understanding
- Speaker verification & identification
- Computer speech recognition
- Machine learning (including Hidden Markov models, neural
networks, Bayesian networks)
- Neurocomputing applications
- Probability and information theory
- Philosphy of mind & neural modelling
- Fitness (including football, skiing, jogging, swimming, squash,
- Painting, drawing, sculpture, architecture, music
(ex guitarist in
- Programming for real-time biometrics based user authentication on
PDA (C++ & Java) (combining voice, face and signature)
- Programming state of the art and novel methods for speaker and
speech recognition (C/C++)
- Applied mathematics (including differential calculus, linear
algebra, information theory, graph theory, operations research,
statistical analysis, machine learning, signal
- HMMs, neural networks & Bayesian networks for speech
recognition, speaker recognition and time series prediction (21 years),
2D computer graphics for map and image processing (5 yrs), 3D computer
graphics for engineering (1 yr).
Associate, Dept. of Phonetics, Saarbrücken University, Germany.
Member of SecurePhone
working on implementation of multi-modal
verification on a PDA (Personal Digital Assistant).
Applied data enhancement technique
from speech recognition to improve performance of speaker recognition.
Research Scientist, Institut Dalle Molle d'Intelligence Artificielle
Perceptive (IDIAP), Martigny,
mathematical model for all-combinations multi-band and multi-stream
Also contributed to other theoretical developments and implementation
of novel techniques for
robust ASR. New techniques were developed in the areas of the
and missing-data processing. Work was within the European Commission
MULTICHAN, COST249, SPHEAR and RESPITE. 2 journal publications. 14
Presented 1 week course "Neural
speech recognition", 18-25 June 2001 Institute for Biomedical
Graz, Austria (for Prof. Gert Pfurtscheller).
Supervised 2 PhD students. Member of PhD jury for Astrid Hagen,
Research associate, Speech and Hearing
Sheffield University, Dept. of Computer Science.
formal solution to the general problem of pattern recognition with
Theoretical and practical development of the "missing data"
noise robust ASR, within the EC SPHERE project. 2 conference
Participated in EC "tempus" project for the
of computer science teaching in Romania.
Research Associate, GTH (Speech
Group), ETSI de Telecomunicación, Universidad Politecnica de
Some work on reducing the computational complexity for a
for simulating the behaviour of small groups of neurons. Applied
model for onset detection to phoneme transition detection in ASR
system. 2 conference publications.
Created system for speech
recognition using only the fraction of speech data which occurs at
phoneme transitions. This was based on a new technique for phoneme
transition detection which was developed during my PhD.
1995 Chairperson, session MOpm2E "Auditory
1995 Member of PhD jury for Sr. Manuel
July 1995, ETSIT, UPM, Madrid.
- 1989-93 Research assistant, European project
"Investigation of automatic speech recognition based on a physiological
model of the auditory system", ICP (Institut de la Communication
INPG, Grenoble, France.
Investigated the role played by cells in
cochlear nucleus in speech recognition. Established (among other things) that type
1 onset neurons are involved in phoneme transition detection, while
type 2 onset cells measure pitch. Developed practical methods for
detecting phoneme transitions and for mapping
distribution of phonetic information over the auditory spectrogram.
Demonstrated that the information theoretical measure of mutual
information is proportional to Pearson's large sample statistic for
testing statistical dependence in contingency tables. 1 jounal
publication. 2 conference publications.
Programmer, Laser-Scan Laboratories Ltd.,
new high speed algorithm (quad-tree processing) for vector image
Contributed to new applications in map and
processing, and geographical information systems, in Fortran and C.
in Alvey project MMI-137, "Real-time 2.5D Vision". 2 conference
Research assistant, computer aided
of Central London, UK.
useful formula concerning the exact shape of the space swept out by a
moving convex polygon - though it was later found that this had
been discovered before!(
Pascal programming applications in 3D
graphics and machine tool control. 2 technical reports.