<?xml version="1.0" encoding="utf-8"?>
<export-example>
  <doc>
    <id>1757</id>
    <completedYear>2020</completedYear>
    <publishedYear>2020</publishedYear>
    <thesisYearAccepted/>
    <language>eng</language>
    <pageFirst/>
    <pageLast/>
    <pageNumber>5</pageNumber>
    <edition/>
    <issue/>
    <volume/>
    <type>conferenceobject</type>
    <publisherName/>
    <publisherPlace/>
    <creatingCorporation/>
    <contributingCorporation/>
    <belongsToBibliography>1</belongsToBibliography>
    <completedDate>--</completedDate>
    <publishedDate>2020-08-11</publishedDate>
    <thesisDateAccepted>--</thesisDateAccepted>
    <title language="eng">Compact Speaker Embedding</title>
    <abstract language="eng">Deep neural networks (DNN) have recently been widely used in speaker recognition systems, achieving state-of-the-art performance on various benchmarks. The x-vector architecture is especially popular in this research community, due to its excellent performance and manageable computational complexity. In this paper, we present the lrx-vector system, which is the low-rank factorized version of the x-vector embedding network. The primary objective of this topology is to further reduce the memory requirement of the speaker recognition system. We discuss the deployment of knowledge distillation for training the lrx-vector system and compare against low-rank factorization with SVD. On the VOiCES 2019 far-field corpus we were able to reduce the weights by 28% compared to the full-rank x-vector system while keeping the recognition rate constant (1.83% EER).</abstract>
    <subTitle language="eng">lrx-vector</subTitle>
    <identifier type="doi">10.48550/arXiv.2008.05011</identifier>
    <identifier type="arxiv">2008.05011</identifier>
    <enrichment key="ConferenceStatement">Annual Conference of the Int’l Speech Communication Association (INTERSPEECH)</enrichment>
    <enrichment key="Reviewstatus">Begutachtet/Reviewed</enrichment>
    <enrichment key="opus.import.date">2024-06-27T13:30:24+00:00</enrichment>
    <enrichment key="opus.source">sword</enrichment>
    <enrichment key="opus.import.user">opus_importer</enrichment>
    <licence>Keine Lizenz - Deutsches Urheberrecht gilt</licence>
    <author>Munir Georges</author>
    <author>Jonathan Huang</author>
    <author>Tobias Bocklet</author>
    <subject>
      <language>deu</language>
      <type>uncontrolled</type>
      <value>speaker recognition, x-vector, low power</value>
    </subject>
    <collection role="institutes" number="">Fakultät Informatik</collection>
  </doc>
</export-example>
