Nara Institute of Science and Technology
Graduate Shcool of Science and Technology, Information Science
2nd-year master's course student
mail: onaka.hien.oj5[at]naist.ac.jp, ohnaka.hien[at]gmail.com
→ This website was migrated to https://i17oonaka-h.github.io/ on September 10, 2025.
This site will no longer be updated.
Spoken dialogue system / 音声対話システム
Turn-taking of human-system dialogue / 人-システム対話のターンテイキング
Audio Signal Processing / 音響信号処理
Speech Enhancement / 音声強調
{Environmental Sound, Speech} Synthesis / {環境音，音声} 合成
Automatic Speech Recognition / 音声認識
Deep Learning / 深層学習
2024/04 - now: NAIST, Graduate School of Sicence and Technology, Information Science (Master's Course) / 奈良先端大 情報科学領域（修士課程）
Laboratory: Intelligent Robot Dialogue lab. (RIKEN) / ロボット対話知能研究室（理化学研究所）
Supervisor: Koichiro Yoshino
2022/04 - 2024/03: NITTC, Advanced Course, Computer Science and Electronic Engineering / 徳山高専 専攻科 情報電子工学専攻
Laboratory: Miyazaki-lab / 宮崎研究室
Supervisor: Ryoichi Miyazaki
2017/04 - 2022/03: NITTC, Regular Course, Computer Science and Electronic Engineering / 徳山高専 情報電子工学科
Laboratory: Miyazaki-lab / 宮崎研究室
Supervisor: Ryoichi Miyazaki
2024/12 - now: LY corp. R&D, (Part-time job)
Mentor: Yuma Shirahata
Theme1: Improving of phonemic and prosodic annotation model for text-to-speech and its sub-task (Accepted to Interspeech2025)
Theme2: work in progress...
2025/02 - 2025/03 (3 weeks): NTT Communication Science Laboratories (Internship)
Theme: Post-refinement of ASR hypotheses for joint multi-channel distant speech recognition
Mentor: Takatomo Kano
2024/08 - 2024/10 (8 weeks): LY corp. R&D, (Internship)
Theme: Improving of phonemic and prosodic annotation model for text-to-speech and its sub-task (Accepted to Interspeech2025)
Mentor: Yuma Shirahata and Ryuichi Yamamoto
2022/06 - 2022/07 (8 weeks): Saruwatari & Koyama lab., The University of Tokyo (Internship)
Theme: Environmental Sound Synthesis from Visual Onomatopoeia (accepted to ICASSP2023)
Mentor: Shinnosuke Takamichi and Keisuke Imoto (Doshisha Univ.)
Journal (Peer-reviewed)
H. Ohnaka and R. Miyazaki, "Unsupervised speech enhancement with spectral kurtosis and double deep priors," Acoustical Science Technology, vol. 46 (2), pp. 146--156, February 2025.
International Conference (Peer-reviewed)
First author
H. Ohnaka, Y. Shirahata, B. Park, and R. Yamamoto, "Grapheme-Coherent Phonemic and Prosodic Annotation of Speech by Implicit and Explicit Grapheme Conditioning," Proc. of Interspeech, August 2025. (to appear)
H. Ohnaka, S. Takamichi, K. Imoto, Y. Okamoto, K. Fujii and H. Saruwatari, "Visual Onoma-to-Wave: Environmental Sound Synthesis from Visual Onomatopoeias and Sound-Source Images," Proc. of International Conference on Acoustics, Speech and Signal Processing (ICASSP), June 2023. [demo, implementation]
Co-author
K. Onishi, H. Ohnaka and K. Yoshino, "Modeling Turn-taking Speed and Speaker Characteristics," Proc. of SigDial, August 2025. (to appear)
国内学会・研究会（査読なし）
大中緋慧，河野誠也，大西一誉，吉野幸一郎，"リアルタイム音声対話システムのための応答タイミングと短文応答の同時予測," 言語処理学会第31回年次大会 (ANLP)，pp. 4063--4067, March 2025 (委員特別賞).
大西一誉，大中緋慧，吉野幸一郎，"話者特性に基づくターンテイキング速度の分析," 言語処理学会第31回年次大会 (ANLP)，pp. 3137--3141, March 2025.
大中緋慧，河野誠也，大西一誉，吉野幸一郎，"情報案内･傾聴対話における対話行為とターン交替時間の関係分析," 言語･音声理解と対話処理研究会 (SIG-SLUD)，pp. 96--101, November 2024.
大中緋慧，宮崎亮一，高道慎之介，"F0に基づいて伸縮された画像文字からの音声合成," 日本音響学会公演論文集，pp.817--820，March 2024. (slide)
中田優翔，広政遼汰，大中緋慧，宮崎亮一，"事前学習不要なDNN音声強調の反復処理に基づく性能改善," 日本音響学会講演論文集，pp.181--184, March 2024.
広政遼汰，大中緋慧，宮崎亮一，"クリーン不要なDNN音声雑音除去に用いる目的信号の分析," 日本音響学会講演論文集，, pp.185--188, March 2024.
中田優翔，広政遼汰，大中緋慧，宮崎亮一，"過学習の抑制に基づく事前学習不要なDNN音声強調," 第26回関西支部若手研究者交流研究発表会，December 2023.
大中緋慧，宮崎亮一，"Double-DIPに基づく教師なし雑音除去の分析と性能改善," 日本音響学会講演論文集，1-Q-2, pp.231--234, March 2023.
大中緋慧，高道慎之介，井本桂右，岡本悠希，藤井一貴，猿渡洋，"Visual onoma-to-wave：音源画像と画像オノマトペを利用した環境音合成の提案," 音声研究会 (SP)，pp.??--??, March 2023. (2024年度 IEEE SPS Tokyo Joint Chapter Student Award)
広政遼汰，大中緋慧，宮崎亮一，"目的信号に雑音を用いるDNN雑音除去の検討," 応用音響研究会 (EA)，pp.??--??, March 2023.
広政遼汰，大中緋慧，宮崎亮一，"クリーン音声を必要としない時間領域でのDNN音声強調," 第25回関西支部若手研究者交流研究発表会，November 2022.
大中緋慧，藤村拓弥，宮崎亮一，"Double-DIP に基づく単一雑音混入音声のみを用いた雑音除去の検討," 応用音響研究会 (EA)，pp.54--61, May 2022.
大中緋慧，藤村拓弥，宮崎亮一，"深層ニューラルネットワークにおける学習過程の信号分布に着目した雑音抑圧の検討," 第24回関西支部若手研究者交流研究発表会，p.45, December 2021.
Others
論文紹介，"Joint Learning of Context and Feedback Embeddings in Spoken Dialogue," pp.61--65, Interspeech読み会2024. (slide)
H. Ohnaka and R. Miyazaki, "Unsupervised speech enhancement with spectral kurtosis and double deep priors," arXiv:2407.03887, 2024.