Publication detail

Improving TRAPS features using LDA

GRÉZL, F., BURGET, L., JAIN, P., ČERNOCKÝ, J.

Original Title

Improving TRAPS features using LDA

English Title

Improving TRAPS features using LDA

Type

conference paper

Language

en

Original Abstract

The goal of this work was to make TRAPS (Temporal Patterns) system smaller and more accurate. Linear Discrimination Analysis (LDA) was used for reducing the dimensionality of temporal vectors. Different ways of processing of these TRAPS-LDA features were tested. Results are promising for further investigation.

English abstract

The goal of this work was to make TRAPS (Temporal Patterns) system smaller and more accurate. Linear Discrimination Analysis (LDA) was used for reducing the dimensionality of temporal vectors. Different ways of processing of these TRAPS-LDA features were tested. Results are promising for further investigation.

Keywords

speech recognition, TRAPs, LDA, temporal patterns

RIV year

2002

Released

16.05.2002

Publisher

Slovak University of Technology in Bratislava

Location

Bratislava

ISBN

80-227-1700-2

Book

Proc. 12th International scientific conference Radioelektronika 2002

Pages count

4

Documents

BibTex


@inproceedings{BUT10263,
  author="František {Grézl} and Lukáš {Burget} and Pratibha {Jain} and Jan {Černocký}",
  title="Improving TRAPS features using LDA",
  annote="The goal of this work was to make TRAPS (Temporal Patterns) system
smaller and more accurate. Linear Discrimination Analysis (LDA) was
used for reducing the dimensionality of temporal vectors. Different
ways of processing of these TRAPS-LDA features were tested. Results are
promising for further investigation.",
  address="Slovak University of Technology in Bratislava",
  booktitle="Proc. 12th International scientific conference Radioelektronika 2002",
  chapter="10263",
  institution="Slovak University of Technology in Bratislava",
  year="2002",
  month="may",
  pages="0",
  publisher="Slovak University of Technology in Bratislava",
  type="conference paper"
}