This page presents the theoretic and practical information about the experiments that have been carried out during this workshop.

PART 1
The Baseline Systems
PART 2
In Search of a Better Starting Point
Looking at WS96 Lexical Pronouciations
Looking at Monophone Durations

Where did TIMIT durations come from?
Where did HTK durations come from?
Where did the ICSI Switchboard system durations come from?
Observations

The experiment to find a better start point
Final Results ... In Short
The System
Training performance on the Cross Validation set
Final Results

Observations

Booting ICSI system from HTK training set alignments
HTK Configuration using the ICSI 56 phone set and RastaPLP12 Features

PART 3
Optimistic or pessimistic

PART 4
The multiband paradigm

Introduction
Approach
Results

4 band experiments
7 band experiments

Theoretical discussion
References

PART 5
Use of syllable-length time-spans in deriving speech feature vectors

The motivation
The cross-correlation evidence
Linear discriminant analysis

PART 6
The speaking rate

PART 7
The modulation spectrum

PART 8
Cohen Hermansky Advanced Features (CHAF)

Result 1 ... CHAF Baseline
Result 2 ... CHAF Baseline + ICSI Baseline no durations
Result 3 ... CHAF Baseline + ICSI Baseline w/ durations
Result 4 ... CHAF + ICSI + 4 Band Multiband
Result 5 ... ICSI Features and CHAF in same MLP

Observations