• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to previous page

Towards a Real-Time Speech Intelligibility Measurement Meter

Research Project

Project/Area Number 17K00223
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Research Field Perceptual information processing
Research InstitutionYamagata University

Principal Investigator

Kondo Kazuhiro  山形大学, 大学院理工学研究科, 教授 (10312753)

Project Period (FY) 2017-04-01 – 2020-03-31
Project Status Completed (Fiscal Year 2019)
Budget Amount *help
¥4,420,000 (Direct Cost: ¥3,400,000、Indirect Cost: ¥1,020,000)
Fiscal Year 2019: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2018: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2017: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
Keywords音声了解度推定 / 加算雑音劣化 / 残響 / 深層学習 / DNN / 音声了解度 / 音声信号処理 / 了解度推定 / 雑音除去 / 残響除去 / 音声情報処理 / 音質推定 / 実時間推定
Outline of Final Research Achievements

We attempted to predict speech intelligibility of speech corrupted with additive noise and reverberation. We predicted clean speech using DNN, and predicted the intelligibility from the difference between predicted and corrupted speech using a separate DNN.
(1) For additive noise, the correlation and RMSE between the predicted and subjective intelligibility was 0.93 and 0.11, respectively.
(2) For reverberation, we first prepared read speech samples of sentences that would be able to measure the effect of reverberation accurately. Using DNN that was trained using the collected speech, the correlation and RMSE between the predicted and subjective intelligibility was 0.94 and 0.06, respectively.

Academic Significance and Societal Importance of the Research Achievements

音声通信等における受話者側の基本的品質尺度の一つに、どの程度正確に聞き取れる音声を伝送できたかを測る音声了解度が挙げられる。音声了解度は複数の聴取者が正確に聞き取れる発話単位数で評価される。一般に多数の聴取者が多数の発話を用いて評価する必要があるため、高価であり時間も多く必要とされる評価となる。このため、受話音声の物理量などのみより音声了解度を推定する。
本研究では、実時間通信でも通話品質が推定できるように、原音を必要としない、メータのように簡単に利用できる音声了解度推定方法の確立を検討した。その結果、深層学習を用いて原音を用いなくても了解度が高精度で推定できることを示した。

Report

(4 results)
  • 2019 Annual Research Report   Final Research Report ( PDF )
  • 2018 Research-status Report
  • 2017 Research-status Report
  • Research Products

    (19 results)

All 2020 2019 2018

All Journal Article (1 results) (of which Peer Reviewed: 1 results) Presentation (18 results) (of which Int'l Joint Research: 6 results)

  • [Journal Article] Estimation of binaural speech intelligibility using machine learning2018

    • Author(s)
      Kondo Kazuhiro、Taira Kazuya
    • Journal Title

      Applied Acoustics

      Volume: 129 Pages: 408-416

    • DOI

      10.1016/j.apacoust.2017.09.001

    • Related Report
      2017 Research-status Report
    • Peer Reviewed
  • [Presentation] 雪の各要因が与える音声了解度への影響 -音素特徴別分析-2020

    • Author(s)
      柴田 崇斗、近藤 和弘
    • Organizer
      日本音響学会春季研究発表会
    • Related Report
      2019 Annual Research Report
  • [Presentation] DNN を用いた残響環境下における音声了解度の推定のための了解度測定方法の検討2020

    • Author(s)
      中澤 和司、近藤 和弘
    • Organizer
      日本音響学会春季研究発表会
    • Related Report
      2019 Annual Research Report
  • [Presentation] 加算雑音劣化音に対するノンレファレンス単語音声了解度推定法の検討 ─未知単語に対する了解度推定精度評価─2020

    • Author(s)
      高橋 宙人、近藤 和弘
    • Organizer
      日本音響学会春季研究発表会
    • Related Report
      2019 Annual Research Report
  • [Presentation] 重回帰分析を用いた雪の各要因が与える音声了解度への影響の分析2019

    • Author(s)
      柴田 崇斗、近藤 和弘
    • Organizer
      電気関係学会東北支部連合大会
    • Related Report
      2019 Annual Research Report
  • [Presentation] 加算雑音劣化音に対するDNNを用いた音声了解度推定と制度評価2019

    • Author(s)
      高橋 宙人、近藤 和弘
    • Organizer
      電気関係学会東北支部連合大会
    • Related Report
      2019 Annual Research Report
  • [Presentation] On Non-Reference Speech Intelligibility Estimation Using DNN Noise Reduction2019

    • Author(s)
      Hiroto Takahashi, Kazuhiro Kondo
    • Organizer
      International Congress on Acoustics
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] De-Reverberation using CNN for Non-Reference Reverberant Speech Intelligibility Estimation2019

    • Author(s)
      Kazushi Nakazawak, Kazuhiro Kondo
    • Organizer
      International Congress on Acoustics
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Multivariate Analysis of Effect of Snow Quality on Speech Intelligibility2019

    • Author(s)
      Shuto Shibata, Kazuhiro Kondo
    • Organizer
      IEEE Global Conference on Consumer Electronics
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] 雪の各要因が与える音声了解度への影響の分析とその補償フィルタの検討2019

    • Author(s)
      柴田 崇斗、近藤 和弘
    • Organizer
      電子情報通信学会電気音響研究会
    • Related Report
      2019 Annual Research Report
  • [Presentation] 残響環境下における音声了解度の評価2019

    • Author(s)
      中澤 和司、近藤 和弘
    • Organizer
      日本音響学会東北支部音響学研究会
    • Related Report
      2019 Annual Research Report
  • [Presentation] 雑音除去を用いたノンレファレンス単語音声了解度推定法の検討2019

    • Author(s)
      高橋 宙人、近藤 和弘
    • Organizer
      日本音響学会春季研究発表会
    • Related Report
      2018 Research-status Report
  • [Presentation] CNNを用いた残響除去の検討2019

    • Author(s)
      中澤 和司、近藤 和弘
    • Organizer
      日本音響学会春季研究発表会
    • Related Report
      2018 Research-status Report
  • [Presentation] Binaural Speech Intelligibility Estimation Using Deep Neural Networks2018

    • Author(s)
      Kondo Kazuhiro, Taira Kazuya, Kobayashi Yosuke
    • Organizer
      Interspeech
    • Related Report
      2018 Research-status Report
    • Int'l Joint Research
  • [Presentation] 加算雑音のある単語音声了解度推定のための深層学習を用いた雑音低減手法の検討2018

    • Author(s)
      高橋宙人、近藤和弘
    • Organizer
      電気関係学会東北支部連合大会
    • Related Report
      2018 Research-status Report
  • [Presentation] On a Reference Signal Estimation from Noisy Speech Using Deep Learning for Intelligibility Estimation2018

    • Author(s)
      Takahashi Hiroto, Kondo Kazuhiro
    • Organizer
      Global Conference on Consumer Electronics
    • Related Report
      2018 Research-status Report
    • Int'l Joint Research
  • [Presentation] De-reverberation Using DNN for Non-reference Reverberant Speech Intelligibility Estimation2018

    • Author(s)
      Nakazawa Kazushi, Kondo Kazuhiro
    • Organizer
      Global Conference on Consumer Electronics
    • Related Report
      2018 Research-status Report
    • Int'l Joint Research
  • [Presentation] 音声了解度推定のための深層学習を用いた雑音除去手法の検討2018

    • Author(s)
      高橋宙人、近藤和弘
    • Organizer
      東北地区若手研究者研究発表会
    • Related Report
      2017 Research-status Report
  • [Presentation] 深層学習を用いた残響の除去の検討2018

    • Author(s)
      中澤和司、近藤和弘
    • Organizer
      東北地区若手研究者研究発表会
    • Related Report
      2017 Research-status Report

URL: 

Published: 2017-04-28   Modified: 2021-02-19  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi