entrance door

Homepage of Yuuki Tachioka

[Japanese] [My CV]

Profile

Education

2002. 4 - 2006. 3Department of Architecture, School of Engineering, The University of Tokyo B(Eng)
2006. 4 - 2008. 3Department of Socio-Cultural Environmental Studies, Graduate School of Frontier Sciences, The University of Tokyo M(Env)
2011. 4 - 2014. 3Department of Japanese Language and Literature, College of Humanities and Sciences, Nihon University B(Literature)
2017. 4 - 2018. 3Department of Information and Communications Engineering, School of Engineering, Tokyo Institute of Technology Dr(Eng)

Work

2008. 4 - 2017.4Mitsubishi Electric Corporation Information Technology R & D center Researcher (Automatic speech recognition)
2012.11 - 2013. 2Mitsubishi Electric Research Laboratories Visitor
2017. 5 -Denso IT Laboratory associate researcher
2018. 4 -Denso IT Laboratory researcher
2022. 4 -Denso IT Laboratory senior researcher

Award

2008. 3Master Thesis Award of Socio-Cultural Environmental Studies, Dept. Socio-Cultural Environmental Studies, Grad. Sch. Frontier Sciences, Univ. Tokyo
2008. 919th Excellent Master's Thesis Prize, Architectural Institute of Japan
2014. 335th Awaya Prize, Acoustical Society of Japan
2018.11ISPACS2018 Best Paper Award
2021. 3IPSJ Yamashita SIG Research Award
2021. 9IPSJ Elevation to senior member
2022.12IEEE GCCE 2022 Excellent Paper Award (Gold prize)

Activity

Information processing society of Japan Senior Member
Acoustical Society of Japan Member
APSIPA chair (2016), Area chair (2023)
Reviewer for journal transactions such as IEEE Trans. ASLP, Speech Communication, and EURASIP Journal and for international conferences such as ICASSP, ASRU, SLT, INTERSPEECH, APSIPA, SCIS&SIS

Research Topic

1. Study on room acoustic analysis(2006-)
2. Study on speech recognition and enhancement(2008-)
3. Metric Japanese analysis on the diary literature in the Heian period(2011-)
4. Spoken dialogue system(2017-)

Publication

Thesis

[abst, dissertation] Robust speech recognition under reverberant and noisy environments, Tokyo Institute of Technology, 2017.
[abst, thesis] Study on the time domain sound field analysis by the CIP method, Graduate School of Frontier Sciences, The University of Tokyo, 2007.
[thesis] Metric study on the diaries in the Heian period -- Study on "Izumishikibu nikki" and "Sarashina Nikki", the College of Humanities and Sciences, Distance Learning Division, The Nihon University 2013.
[abst, thesis] Geometrical acoustic simulation using scattering coefficients to predict the reverberation time of non-diffuse sound fields, Department of Architecture, School of Engineering, The University of Tokyo, 2005.

Book

[Z01] Y. Tachioka, T. Hanazawa, T. Narita, and J. Ishii: Advanced ASR Technologies for Mitsubishi Electric Speech Applications in "New Era for Robust Speech Recognition" Eds. S. Watanabe, M. Delcroix, F. Metze, and J. R Hershey, pp.419-429, Springer, 2017

Selected Refreed Paper (All papers are listed in the Japanese page

[A15] Y. Tachioka, S. Watanabe, J. Le Roux, and J. R. Hershey: Prior-based binary masking and discriminative methods for reverberant and noisy speech recognition using distant stereo microphones, Journal of Information Processing vol.25 no.6, pp.407-416, 2017. 6.
[A14] Y. Tachioka and T. Narita: Template-based method for compensation of time difference of arrival in passive sound source localization under reverberant and noisy environments, Journal of Signal Processing vol.21 no.2, pp.73-79, 2017. 3.
[A12] Y. Tachioka and J. Ishii: Long short-term memory recurrent-neural-network-based bandwidth extension for automatic speech recognition, Acoustical Science & Technology vol.37 no.6, pp.319-321, 2016. 11.
[A10] Y. Tachioka, T. Narita, and S. Watanabe: Effectiveness of dereverberation, feature transformation, discriminative training methods, and system combination approach for various reverberant environments, EURASIP Journal on Advances in Signal Processing, 2015:52 doi:10.1186/s13634-015-0241-y, 2015. 6.
[A09] Y. Tachioka, T. Narita, and J. Ishii: Semi-blind source separation using binary masking and indepedent vector analysis, IEEJ Transactions on Electrical and Electronic Engineering vol.10 no.1, pp.114-115, 2015. 1.
[A08] Y. Tachioka, T. Narita, and J. Ishii: Estimation of speech recognition performance for clipped speech based on objective measures, Acoustical Science and Technology vol.35 no.6, pp.324-326, 2014. 11.
[A05] Y. Tachioka, T. Hirano, and J. Ishii: Elimination of artifacts on aperture synthesis images using ultrasonic sensor's directivity and the ratio of intensity, IEEJ Transactions on Electrical and Electronic Engineering vol.8 no.4, pp., 2013. 7.
[A04] Y. Tachioka, T. Hanazawa, and T. Iwasaki: Dereverberation method with reverberation time estimation using floored ratio of spectral subtraction, Acoustical Science & Technology vol.34 no.3, pp.212-215, 2013. 5.
[A03] Y. Tachioka: Boundary condition for finite-difference time-domain method using digital filters and efficient design of filter coefficients using equivalent mechanical system, Acoustical Science & Technology vol.33 no.5, pp.329-331, 2012. 9.
[A02] Y. Tachioka, T. Narita, and T. Iwasaki: Direction of arrival estimation by cross-power spectrum phase analysis using prior distributions and voice activity detection information, Acoustical Science & Technology vol.33 no.1, pp.68-71, 2012. 1.
[A01] Y. Tachioka, Y. Yasuda, and T. Sakuma: Application of the constrained interpolation profile method to room acoustic problems--Examination of boundary modeling and spatial/time discretization, Acoustical Science & Technology vol.33 no.1, pp.21-32, 2012. 1.

International Conference

[B46] Y. Tachioka: ditlab at the NTCIR-17 QA Lab-PoliInfo-4 Task, NTCIR 17 Conference: Proceedings of the 17th NTCIR Conference on Evaluation of Information Access Technologies, NII, pp. 241-243, 2023. 12. [refereed]
[B45] Y. Tachioka: ditlab at the NTCIR-17 Transfer Task, NTCIR 17 Conference: Proceedings of the 17th NTCIR Conference on Evaluation of Information Access Technologies, NII, pp. 330-331, 2023. 12. [refereed]
[B44, pdf] Y. Tachioka: Analysis of embedding space of speech synthesis system as high-dimension, low-sample-size data, IEEE 12th Global Conference on Consumer Electronics (GCCE 2023), Nara Royal Hotel, pp.443-446, 2023. 10. [refereed(289+141(poster)/650; 68%)]
[B43] Y. Tachioka: Conditioning of variational autoencoder by user traits for item recommendation, Music recommender systems workshop @ 17th ACM Conference on Recommender Systems (RecSys 2023), Singapore, Suntec city, pp.1-7, 2023. 9. [refereed(7/15; 47%)]
[B42] Y. Tachioka: Conditioning of autoencoder for various types of anomaly sound detection by using single model, The 11th International Conference on Computer and Communications Management (ICCCM) 2023, Nagoya Institute of Technology, 2023. 8. [refereed]
[B41] Y. Tachioka: Impulse response classification based on persistent homology, TDA week 2023, Kyoto university, 2023. 8. [refereed]
[B40] Y. Tachioka: Privacy preservation satisfying utility requirements based on multi-objective optimization, Joint 12th International Conference on Soft Computing and Intelligent Systems and 23rd International Symposium on Advanced Intelligent Systems (SCIS&ISIS), Ise-Shima, pp.1-4, 2022. 11. [refereed(72/105; 68.6%)]
[B39] Y. Tachioka: Impulse response classification focusing on geometrical structures, The 24th International Congress on Acoustics (ICA), Gyeongju, pp.1-7, 2022. 10.
[B38] Y. Tachioka: Multi-condition training and system combination for automatic MOS prediction, IEEE 11th Global Conference on Consumer Electronics (GCCE 2022), Senri Life Science Center, pp.353-354, 2022. 10. [refereed(231(oral)+102(poster)/502; 68.1%)] IEEE GCCE 2022 Excellent Paper Award (Gold prize)
[B37] Y. Tachioka and A. Keyaki: ditlab at the NTCIR-16 QA Lab-PoliInfo-3, NTCIR 16 Conference: Proceedings of the 16th NTCIR Conference on Evaluation of Information Access Technologies, Online, pp. 207-210, 2022. 6. [refereed]
[B36] Y. Tachioka: Application of cross-data matrix methodology to embedded vector of deep-learning models with estimation of spiked model parameter, The 11th Conference of the IASC-ARS (The Asian Regional Section of the International Association for Statistical Computing), Doshisha Univ., 2022. 2. [refereed]
[B35] Y. Tachioka: Integration of annotator-wise estimations for emotion recognition by using group softmax, The 13th Asia-Pacific Signal and Information Processing Association (APSIPA) Annual Summit and Conference, Online, pp. 694-699, 2021. 12. [refereed(191/283; 67.5%)]
[B34] K. Ozawa, S. Sumiyoshi, and Y. Tachioka: Snapshot multispectral image completion and unmixing with total variation regularization on abundance maps, The 13th Asia-Pacific Signal and Information Processing Association (APSIPA) Annual Summit and Conference, Online, pp. 1367-1374, 2021. 12. [refereed(191/283; 67.5%)]
[B33] Y. Tachioka: Exponentiated geometric-mean-balanced mask for underdetermined source separation, IEEE 10th Global Conference on Consumer Electronics (GCCE 2021), Mielparque Kyoto+Online, pp.344-345, 2021. 10. [refereed(257(oral)+93(poster)/525; 69.7%)]
[B32] Y. Tachioka: Privacy preserving acoustic model training for speech recognition, Asia-Pacific Signal and Information Processing Association (APSIPA) Annual Summit and Conference, Online, pp. 627-631, 2020. 12. [refereed(204/265; 71.6%)]
[B31, pdf] Y. Tachioka: Multilingual acoustic model training based on phoneme set conversion, IEEE 9th Global Conference on Consumer Electronics (GCCE 2020), Kobe, pp.458-459, 2020. 10. [refereed(241(oral)+89(poster)/473; 71%)]
[B30, pdf] T. Ushijima, Y. Tachioka, S. Uenohara, and K. Furuya: Sparse independent vector analysis based on Mel filter, IEEE 9th Global Conference on Consumer Electronics (GCCE 2020), Kobe, pp.826-827, 2020. 10. [refereed(241(oral)+89(poster)/473; 71%)]
[B29] T. Izumi, Y. Tachioka, S. Uenohara, and K. Furuya: Activation driven synchronized joint diagonalization for underdetermined sound source separation, Asia-Pacific Signal and Information Processing Association (APSIPA) Annual Summit and Conference, Gansu International Conference Center, pp. 1318-1322, 2019. 11. [refereed]
[B28] Y. Tachioka: Hypothesis correction based on semi-character recurrent neural network for end-to-end speech recognition, Asia-Pacific Signal and Information Processing Association (APSIPA) Annual Summit and Conference, Gansu International Conference Center, pp. 862-867, 2019. 11. [refereed(187/272; 68.8%)]
[B27] Y. Tachioka: Knowledge distillation using soft and hard labels and annealing for acoustic model training, IEEE 8th Global Conference on Consumer Electronics (GCCE 2019), Senri Life Science Center, pp.715-716, 2019. 10. [refereed(262(oral)+158(poster)/666; 65.3%)]
[B26] Y. Tachioka: Permutation alignment based on MUSIC spectrum discrepancy for blind source separation, The 27th European Signal Processing Conference (EUSIPCO), PALEXCO, A Coruna, pp.1958-1962, 2019. 9. [refereed(500/820)]
[B25] T. Izumi, Y. Tachioka, S. Uenohara, K. Furuya: Automatic determination of the optimum number of updates in synchronized joint diagonalization, The 11th International Workshop on Virtual Environment and Network-Oriented Applications (VENOA-2019), Sydney, 2019. 7. [refereed]
[B24] Y. Tachioka: Multi-channel non-negative matrix factorization initialized with full-rank and rank-1 spatial correlation matrix for speech recognition, The 26th International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS), Ishigaki Resort ANA Continental Hotel, pp.246-251, 2018. 11. ISPACS2018 Best paper award [refereed(99/134; 72%)]
[B23, pdf] Y. Tachioka: DNN-based voice activity detection using auxiliary speech models in noisy environments, The 43th International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Calgary, pp.5529-5533, 2018. 4.[refereed(1406/2830; 49.7%)]
[B22, pdf] T. Uramoto, Y. Tachioka, T. Narita, I. Miura, S. Uenohara, K. Furuya: Sequential initialization of multichannel nonnegative matrix factorization for sound source separation, IEEE 6th Global Conference on Consumer Electronics (GCCE 2017), WINC AICHI, DOI: 10.1109/GCCE.2017.8229207, 2017. 10.[refereed]
[B21] Y. Tachioka, T. Narita, I. Miura, T. Uramoto, N. Monta, S. Uenohara, K. Furuya, S. Watanabe, and J. Le Roux: Coupled initialization of multi-channel non-negative matrix factorization based on spatial and spectral information, The 18th Annual Conference of the International Speech Communication Association (INTERSPEECH), Stockholm, pp.2461-2465, 2017. 8.[refereed(799/1582; 52%)]
[B20, pdf] Y. Tachioka and T. Narita: Optimal automatic speech recognition system selection for noisy environments, Asia-Pacific Signal and Information Processing Association (APSIPA) Annual Summit and Conference, Jeju, pp.1-8, 2016. 12.[refereed(242 accepted; 78%)]
[B19] I. Miura, Y. Tachioka, T. Narita, J. Ishii, F. Yoshiyama, S. Uenohara, and K. Furuya: Multi-channel non-negative matrix factorization with binary mask initialization for automatic speech recognition, 5th Joint Meeting ASA/ASJ, Honolulu, 2016. 12.
[B18] Y. Tachioka, S. Watanabe, and T. Hori: The MELCO/MERL system combination approach for the fourth CHiME challenge, The fourth CHiME challenge workshop, San Francisco, pp.1-3, 2016. 9.
[B17] H. Kanagawa, Y. Tachioka, S. Watanabe, and Jun Ishii: Feature-space structural MAPLR with regression tree-based multiple transformation matrices for DNN, Asia-Pacific Signal and Information Processing Association (APSIPA) Annual Summit and Conference, Shanghai, pp.86-92, 2015. 12.[refereed(266 accepted; 69%)]
[B16] Y. Tachioka and S. Watanabe: Uncertainty training and decoding methods of deep neural networks based on stochastic representation of enhanced features, The 16th Annual Conference of the International Speech Communication Association (INTERSPEECH), Dresden, pp.3541-3545, 2015. 9.[refereed(746/1458; 51%)]
[B15] Y. Tachioka and S. Watanabe: Discriminative method for recurrent neural network language models, The 40th International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Brisbane, pp.5386-5390, 2015. 4.[refereed(1207/2322; 52%)]
[B14] Y. Tachioka, S. Watanabe, J. Le Roux, and J. R. Hershey: Sequential discriminative training for low-rank deep neural network, The 2nd IEEE Global Conference on Signal and Information Processing (GlobalSIP), Atlanta, pp.735-739, 2014. 12.[refereed]
[B13] Y. Tachioka, S. Watanabe, J. Le Roux, and J. R. Hershey: Sequential maximum mutual information linear discriminant analysis for speech recognition, The 15th Annual Conference of the International Speech Communication Association (INTERSPEECH), Singapore, pp.2415-2419, 2014. 9.[refereed(614/1173; 52%)]
[B12] Y. Tachioka, T. Narita, S. Watanabe, and J. Le Roux: Ensamble integration of calibrated speaker localization and statistical speech detection, The 4th workshop on Hands-free Speech Communication and Microphone Array (HSCMA), Nancy, ID 1569901761 pp.1-5, 2014. 5.[refereed(41/55)]
[B11] Y. Tachioka, T. Narita, S. Watanabe, and F. Weninger: Dual system combination approach for various reverberant environments, REVERB challenge, Florence, ID 1569886337 pp.1-8, 2014. 5.[refereed]
[B10] F. Weninger, S. Watanabe, J. Le Roux, J. Hershey, Y. Tachioka J. T Geiger, B. W Schuller, and G. Rigoll: The MERL/MELCO/TUM system using deep recurrent neural network speech enhancement, REVERB challenge, Florence, ID 1569884459 pp.1-8, 2014. 5.[refereed]
[B09] F. Weninger, S. Watanabe, Y. Tachioka, and B. Schuller: Deep recurrent de-noising auto-encoder and blind de-reverberation for reverberated speech recognition, The 39th International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Florence, pp.4656-4660, 2014. 5.[refereed(1709/3500)]
[B08] Y. Tachioka and S. Watanabe: A generalized framework of discriminative training for system combination, Automatic Speech Recognition and Understanding Workshop (ASRU), Olomouc, pp.43-48, 2013. 12.[refereed]
[B07] Y. Tachioka, T. Narita, T. Hanazawa, and J. Ishii: Voice activity detection based on density ratio estimation and system combination, Asia-Pacific Signal and Information Processing Association (APSIPA) Annual Summit and Conference, Kaohsiung, pp.1-4, 2013. 11.[refereed]
[B06] Y. Tachioka: Objective measurement of variants in classical literature, The International Conference on Culture and Computing (Culture and Computing 2013), Kyoto, pp.202-203, 2013. 9.[refereed]
[B05] Y. Tachioka and S. Watanabe: Discriminative training of acoustic models for system combination, 14th Annual Conference of the International Speech Communication Association (INTERSPEECH), Lyon, pp.2355-2359, 2013. 8.[refereed]
[B04] Y. Tachioka: Optimal design of sound reflectors by particle swarm optimization, The 21st International Congress on Acoustics (ICA), Montreal, pp.1-9, 2013. 6.
[B03] Y. Tachioka, S. Watanabe, J. Le Roux, and J. R. Hershey: Discriminative methods for noise robust speech recognition: A CHiME Challenge Benchmark, The 2nd International Workshop on Machine Listening in Multisource Environments, Vancouver, pp.19-24, 2013. 6.[refereed]
[B02] Y. Tachioka, S. Watanabe, and J. R. Hershey: Effectiveness of discriminative training and feature transformation for reverberated and noisy speech, The 38th International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Vancouver, pp.6935-6939, 2013. 5.[refereed]
[B01] T. Sakuma, Y. Kosaka and Y. Tachioka: Numerical determination of scattering coefficients of wall surfaces for geometrical room acoustic simulation, 4th Joint Meet. ASA/ASJ (Honolulu), J. Acoust. Soc. Am. Vol.120, pp.2998, 2006. [invited]

Talk or seminar

[D3] Y. Tachioka: Discriminative methods for noise robust speech recognition: A CHiME Challenge Benchmark, Johns Hopkins University talk, 2013. 6.
[D2] Y. Tachioka: Discriminative methods for noise robust speech recognition: A CHiME Challenge Benchmark, MIT sls seminar, 2013. 6.
[D1] Y. Tachioka: Introduction to reseach labs by young researchers, "Information Tech R&D center, Mitsubishi Electric", SLP workshop, 2012. 7.

Created3/Apr/13
Updated5/Oct/23