Enhancing LSTM rnn-based speech overlap detection by artificially mixed data

Gerhard Hagerer; Vedhas Pandit; Florian Eyben; Bjorn Schuller

Enhancing LSTM rnn-based speech overlap detection by artificially mixed data

Gerhard Hagerer, Vedhas Pandit, Florian Eyben, Bjorn Schuller

Research output: Contribution to conference › Paper › peer-review

14 Scopus citations

Abstract

This paper presents a new method for Long Short-Term Memory Recurrent Neural Network (LSTM) based speech overlap detection. To this end, speech overlap data is created artificially by mixing large amounts of speech utterances. Our elaborate training strategies and presented network structures demonstrate performance surpassing the considered state-of-the-art overlap detectors. Thereby we target the full ternary task of non-speech, speech, and overlap detection. Furthermore, speakers' gender is recognised, as the first successful combination of this kind within one model.

Original language	English
Pages	45-52
Number of pages	8
State	Published - 2017
Externally published	Yes
Event	3rd AES International Conference on Semantic Audio 2017 - Erlangen, Germany Duration: 22 Jun 2017 → 24 Jun 2017

Conference

Conference	3rd AES International Conference on Semantic Audio 2017
Country/Territory	Germany
City	Erlangen
Period	22/06/17 → 24/06/17

Cite this

@conference{252d83a7f03c46a9ac67fa5a2f57db78,

title = "Enhancing LSTM rnn-based speech overlap detection by artificially mixed data",

abstract = "This paper presents a new method for Long Short-Term Memory Recurrent Neural Network (LSTM) based speech overlap detection. To this end, speech overlap data is created artificially by mixing large amounts of speech utterances. Our elaborate training strategies and presented network structures demonstrate performance surpassing the considered state-of-the-art overlap detectors. Thereby we target the full ternary task of non-speech, speech, and overlap detection. Furthermore, speakers' gender is recognised, as the first successful combination of this kind within one model.",

author = "Gerhard Hagerer and Vedhas Pandit and Florian Eyben and Bjorn Schuller",

year = "2017",

language = "English",

pages = "45--52",

note = "3rd AES International Conference on Semantic Audio 2017 ; Conference date: 22-06-2017 Through 24-06-2017",

}

TY - CONF

T1 - Enhancing LSTM rnn-based speech overlap detection by artificially mixed data

AU - Hagerer, Gerhard

AU - Pandit, Vedhas

AU - Eyben, Florian

AU - Schuller, Bjorn

PY - 2017

Y1 - 2017

N2 - This paper presents a new method for Long Short-Term Memory Recurrent Neural Network (LSTM) based speech overlap detection. To this end, speech overlap data is created artificially by mixing large amounts of speech utterances. Our elaborate training strategies and presented network structures demonstrate performance surpassing the considered state-of-the-art overlap detectors. Thereby we target the full ternary task of non-speech, speech, and overlap detection. Furthermore, speakers' gender is recognised, as the first successful combination of this kind within one model.

AB - This paper presents a new method for Long Short-Term Memory Recurrent Neural Network (LSTM) based speech overlap detection. To this end, speech overlap data is created artificially by mixing large amounts of speech utterances. Our elaborate training strategies and presented network structures demonstrate performance surpassing the considered state-of-the-art overlap detectors. Thereby we target the full ternary task of non-speech, speech, and overlap detection. Furthermore, speakers' gender is recognised, as the first successful combination of this kind within one model.

UR - http://www.scopus.com/inward/record.url?scp=85028085834&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:85028085834

SP - 45

EP - 52

T2 - 3rd AES International Conference on Semantic Audio 2017

Y2 - 22 June 2017 through 24 June 2017

ER -

Enhancing LSTM rnn-based speech overlap detection by artificially mixed data

Abstract

Conference

Other files and links

Fingerprint

Cite this