Improving keyword spotting with a tandem BLSTM-DBN architecture

Martin Wöllmer, Florian Eyben, Alex Graves, Björn Schuller, Gerhard Rigoll

Publikation: Beitrag in Buch/Bericht/KonferenzbandKonferenzbeitragBegutachtung

4 Zitate (Scopus)

Abstract

We propose a novel architecture for keyword spotting which is composed of a Dynamic Bayesian Network (DBN) and a bidirectional Long Short-Term Memory (BLSTM) recurrent neural net. The DBN uses a hidden garbage variable as well as the concept of switching parents to discriminate between keywords and arbitrary speech. Contextual information is incorporated by a BLSTM network, providing a discrete phoneme prediction feature for the DBN. Together with continuous acoustic features, the discrete BLSTM output is processed by the DBN which detects keywords. Due to the flexible design of our Tandem BLSTM-DBN recognizer, new keywords can be added to the vocabulary without having to re-train the model. Further, our concept does not require the training of an explicit garbage model. Experiments on the TIMIT corpus show that incorporating a BLSTM network into the DBN architecture can increase true positive rates by up to 10%.

OriginalspracheEnglisch
TitelAdvances in Nonlinear Speech Processing - International Conference on Nonlinear Speech Processing, NOLISP 2009, Revised Selected Papers
Seiten68-75
Seitenumfang8
DOIs
PublikationsstatusVeröffentlicht - 2010
VeranstaltungInternational Conference on Nonlinear Speech Processing, NOLISP 2009 - Vic, Spanien
Dauer: 25 Juni 200927 Juni 2009

Publikationsreihe

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Band5933 LNAI
ISSN (Print)0302-9743
ISSN (elektronisch)1611-3349

Konferenz

KonferenzInternational Conference on Nonlinear Speech Processing, NOLISP 2009
Land/GebietSpanien
OrtVic
Zeitraum25/06/0927/06/09

Fingerprint

Untersuchen Sie die Forschungsthemen von „Improving keyword spotting with a tandem BLSTM-DBN architecture“. Zusammen bilden sie einen einzigartigen Fingerprint.

Dieses zitieren