Kazuhito Koishida

I am a Principal Lead Scientist at Applied Sciences Group in Experiences + Devices organization. I have been with Microsoft since 2000. My area of interests is in signal processing and machine learning for audio, speech, computer vision, and other sensor data.

Past projects

• Audio and voice compression: Bitrate/bandwidth scalable codec, MELP codec at 1.2kbps, and Windows Media Audio and Voice codec
• Audio matching: Voice note application and music recognition service
• Microphone array processing: Beamforming and sound source localization
• Audio/voice detection and recognition: Keyword spotting and speaker identification
• Speech enhancement: Audio/visual fusion and bandwidth expansion

Education

• B.S degree in Electrical Engineering from the Tokyo Institute of Technology, Japan, in 1994
• M.S. degree in Electrical Engineering from the Tokyo Institute of Technology, Japan, in 1995
• Ph.D. degree in Electrical Engineering from the Tokyo Institute of Technology, Japan, in 1998. Dissertation title: Speech Coding Based on Mel-Generalized Cepstral Analysis
• Post doctoral researcher at Signal Compression Lab in the University of California, Santa Barbara, 1998-2000

• CorrGAN: Simultaneous Learning of Speech Enhancement and Perceptual Quality Loss Functions

  To appear in ICASSP 2025 2025

  By Vasily Zadorozhnyy, Saeed Amizadeh, Qiang Ye, Kazuhito Koishida
• Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale

  NeurIPS Workshop 2024 December, 2024

  By Rogerio Bonatti, Danny Zhao, Francesco Bonacci, Dillon Dupont, Sara Abdali, Yinheng Li, Justin Wagle, Kazuhito Koishida, Arthur Bucker, Lawrence Jang, Zack Hui
• VideoWebArena: Evaluating Long Context Multimodal Agents with Video Understanding Web Tasks

  NeurIPS Workshop 2024 December, 2024

  By Lawrence Jang, Yinheng Li, Charles Ding, Justin Lin, Paul Pu Liang, Danny Zhao, Rogerio Bonatti, Kazuhito Koishida
• Learned Image Compression with Text Quality Enhancement

  Proceedings, 2024 IEEE International Conference on Image Processing (ICIP) October, 2024

  By Andrew Lai (Chih-Yu Lai), Dung Tran, Kazuhito Koishida
• Automatic Disfluency Detection From Untranscribed Speech

  IEEE/ACM Transactions on Audio, Speech, and Language Processing October, 2024 Vol. 32 Pages 4727-4740

  By Amrit Romana, Kazuhito Koishida, Emily Mower Provost
• ConsistencyTTA: Accelerating Diffusion-Based Text-to-Audio Generation with Consistency Distillation

  Proceedings, Interspeech 2024 September, 2024

  By Yatong Bai, Trung Dang, Dung Tran, Kazuhito Koishida, Somayeh Sojoudi
• LiveSpeech: Low-Latency Zero-Shot Text-to-Speech via Autoregressive Modeling of Audio Discrete Codes

  Proceedings, Interspeech 2024 September, 2024

  By Trung Dang, David Aponte, Dung Tran, Kazuhito Koishida
• Data Generation using Large Language Models for Text Classification: An Empirical Case Study

  DMLR Workshop in ICML 2024 July, 2024

  By Yinheng Li, Rogerio Bonatti, Sara Abdali, Justin Wagle, Kazuhito Koishida
• Weakly-Supervised Audio Separation via Bi-modal Semantic Similarity

  Proceedings of the Twelfth International Conference on Learning Representations (ICLR) May, 2024 Vol. abs/2404.01740

  By Tanvir Mahmud, Saeed Amizadeh, Kazuhito Koishida, Diana Marculescu
• uaMix-MAE: Efficient Tuning of Pretrained Audio Transformers with Unsupervised Audio Mixtures

  Proceedings. ICASSP 2024 April, 2024 Pages 5435-5439

  By Afrina Tabassum, Dung Tran, Trung Dang, Ismini Lourentzou, Kazuhito Koishida
• Improving Low-Latency Mono-Channel Speech Enhancement by Compensation Windows in STFT Analysis

  Proceedings. International Conference on Complex Networks and Their Applications November, 2023 Pages 363-373 ISBN: 978-3-031-53468-3

  By Minh Bui, Dung Tran, Kazuhito Koishida, Trac D. Tran, Peter Chin
• SCP-GAN: Self-Correcting Discriminator Optimization for Training Consistency Preserving Metric GAN on Speech Enhancement Tasks

  Proceedings, Interspeech 2023 August, 2023 Pages 2463-2467

  By Vasily Zadorozhnyy, Qiang Ye, Kazuhito Koishida
• Progressive Knowledge Distillation: Balancing Inference Latency and Accuracy at Runtime

  Workshop on Efficient Systems for Foundation Models @ ICML2023 July, 2023

  By Don Dennis, Abhishek Shetty, Anish Sevekari, Kazuhito Koishida, Virginia Smith
• Toward A Multimodal Approach for Disfluency Detection and Categorization

  Proceedings, ICASSP 2023 June, 2023 Pages 1-5

  By Amrit Romana, Kazuhito Koishida
• Training Robust Zero-Shot Voice Conversion Models with Self-supervised Features

  Proceedings, ICASSP 2022 May, 2022 Pages 6557-6561

  By Trung Dang, Dung Tran, Peter Chin, Kazuhito Koishida
• A Training Framework for Stereo-Aware Speech Enhancement Using Deep Neural Networks

  Proceedings, ICASSP 2022 May, 2022 Pages 6962-6966

  By Bahareh Tolooshams, Kazuhito Koishida
• Single-Channel Speech Enhancement Using Learnable Loss Mixup

  Proceedings, Interspeech 2021 August, 2021 Pages 2696-2700

  By Oscar Chang, Dung Tran, Kazuhito Koishida
• INTERSPEECH 2021 Deep Noise Suppression Challenge

  Proceedings, Interspeech 2021 August, 2021 Pages 2796-2800

  By Chandan K A Reddy, Harishchandra Dubey, Kazuhito Koishida, Arun Nair, Vishak Gopal, Ross Cutler, Sebastian Braun, Hannes Gamper, Robert Aichner, Sriram Srinivasan
• Cascaded Time + Time-Frequency Unet for Speech Enhancement: Jointly Addressing Clipping, Codec Distortions, and Gaps

  Proceedings, ICASSP 2021 June, 2021 Pages 7153-7157

  By Arun Nair, Kazuhito Koishida
• Online Directional Speech Enhancement Using Geometrically Constrained Independent Vector Analysis

  Proceedings, Interspeech 2020 October, 2020 Pages 61-65

  By Li Li, Kazuhito Koishida, Shoji Makino
• Low-Latency Single Channel Speech Dereverberation using U-Net Convolutional Neural Networks

  Proceedings, Interspeech 2020 October, 2020 Pages 2442-2446

  By Ahmet Bulut, Kazuhito Koishida
• Robust Pitch Regression with Voiced/Unvoiced Classification in Nonstationary Noise Environments

  Proceedings, Interspeech 2020 October, 2020 Pages 175-179

  By Dung Tran, Uros Batricevic, Kazuhito Koishida
• Single-Channel Speech Enhancement by Subspace Affinity Minimization

  Proceedings, Interspeech 2020 October, 2020 Pages 2447-2451

  By Dung Tran, Kazuhito Koishida
• Neuro-Symbolic Visual Reasoning: Disentangling "Visual" from "Reasoning"

  Proceedings of the 37th International Conference on Machine Learning (ICML), Vienna, Austria July, 2020

  By Saeed Amizadeh, Hamid Palangi, Oleksandr Polozov, Yichen Huang, Kazuhito Koishida
• MMTM: Multimodal Transfer Module for CNN Fusion

  The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) June, 2020

  By Hamid Vaezi Joze, Amirreza Shaban, Michael Iuzzolino, Kazuhito Koishida
• Improved Active Speaker Detection Based on Optical Flow

  IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) June, 2020 Pages 4084-4090

  By Chong Huang, Kazuhito Koishida
• Low-Latency Single Channel Speech Enhancement Using U-Net Convolutional Neural Networks

  Proceedings, ICASSP 2020 May, 2020 Pages 6214-6218

  By Ahmet Bulut, Kazuhito Koishida
• Geometrically Constrained Independent Vector Analysis For Directional Speech Enhancement

  Proceedings, ICASSP 2020 May, 2020 Pages 846-850

  By Li Li, Kazuhito Koishida
• AV(SE)2 : Audio-Visual Squeeze-Excite Speech Enhancement

  Proceedings, ICASSP 2020 May, 2020 Pages 7539-7543

  By Michael Iuzzolino, Kazuhito Koishida
• Sound Event Detection in Multichannel Audio Using Convolutional Time-Frequency-Channel Squeeze and Excitation

  Proceedings, Interspeech 2019 September, 2019 Pages 3629-3633

  By Wei Xia, Kazuhito Koishida
• Adversarial Training for Speech Super-Resolution

  IEEE Journal of Selected Topics in Signal Processing May, 2019 Vol. 13, No. 2 Pages 347-358

  By Sefik Emre Eskimez, Kazuhito Koishida, Zhiyao Duan
• Speech Super Resolution Generative Adversarial Network

  Proceedings, 2019 International Conference on Acoustics, Speech and Signal Processing (ICASSP) May, 2019 Pages 3717-3721

  By Sefik Emre Eskimez, Kazuhito Koishida
• Text Independent Speaker Verification Based on Triplet Convolutional Neural Network Embeddings

  IEEE/ACM Transactions on Audio, Speech, and Language Processing September, 2018 Vol. 26, No. 9 Pages 1633-1644

  By Chunlei Zhang, Kazuhito Koishida, John H. L. Hansen
• End-to-End Text-Independent Speaker Verification with Flexibility in Utterance Duration

  Proceedings, 2017 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU) December, 2017 Pages 584-590

  By Chunlei Zhang, Kazuhito Koishida
• End-to-End Text-Independent Speaker Verification with Triplet Loss on Short Utterances

  Proceedings, Interspeech 2017 2017 Pages 1487-1491

  By Chunlei Zhang, Kazuhito Koishida
• Hybrid Low Bitrate Audio Coding Using Adaptive Gain Shape Vector Quantization

  Proceedings, 2008 IEEE 10th Workshop on Multimedia Signal Processing October, 2008 Pages 927-932

  By Sanjeev Mehrotra, Wei-ge Chen, Kazuhito Koishida, Naveen Thumpudi
• A 1200/2400 BPS Coding Suite Based on MELP

  Speech Coding, 2002, IEEE Workshop Proceedings October, 2002 Pages 90-92

  By Tian Wang, Kazuhito Koishida, Vladimir Cuperman, Allen Gersho, J.S. Collura
• Vector Quantization of Speech Spectral Parameters Using Statistics of Static and Dynamic Features

  IEICE Transactions on Information and Systems October, 2001 Vol. E84-D, No. 10 Pages 1427-1434

  By Kazuhito Koishida, Keiichi Tokuda, Takashi Masuko, Takao Kobayashi
• Enhancing MPEG-4 CELP by Jointly Optimized Inter/Intra-frame LSP Predictors

  2000 IEEE Workshop on Speech Coding. Proceedings. Meeting the Challenges of the New Millennium (Cat. No.00EX421) September, 2000 Pages 90-92

  By Kazuhito Koishida, Jan Lindén, Vladimir Cuperman, Allen Gersho
• A 16-kbit/s Bandwidth Scalable Audio Coder Based on the G.729 Standard

  Proceedings, 2000 International Conference on Acoustics, Speech and Signal Processing (ICASSP) June, 2000 Pages 1149-1152

  By Kazuhito Koishida, Vladimir Cuperman, Allen Gersho
• A 1200 BPS Speech Coder Based on MELP

  2000 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.00CH37100) June, 2000 Vol. 3 Pages 1375-1378

  By Tian Wang, Kazuhito Koishida, Vladimir Cuperman, Allen Gersho, J.S. Collura
• A 16 kb/s Wideband CELP-based Speech Coder Using Mel-Generalized Cepstral Analysis

  IEICE Transactions on Information and Systems April, 2000 Vol. 83 Pages 876-883

  By Kazuhito Koishida, Gou Hirabayashi, Keiichi Tokuda, Takao Kobayashi
• A Wideband CELP Speech Coder at 16 kbit/s Based on Mel-Generalized Cepstral Analysis

  Proceedings of the 1998 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '98 (Cat. No.98CH36181) May, 1998 Vol. 1 Pages 161-164

  By Kazuhito Koishida, Keiichi Tokuda, Gou Hirabayashi, Takao Kobayashi
• CELP Speech Coding Based on Mel-Generalized Cepstral Analysis

  IEICE Transactions on Information and Systems February, 1998 Vol. J81-A, No. 2 Pages 252-260

  By Kazuhito Koishida, Keiichi Tokuda, Satoshi Imai, Takao Kobayashi
• A 16 kbit/s Wideband CELP Coder Using Mel-Generalized Cepstral Analysis and Its Subjective Evaluation

  Proceedings, 5th International Conference on Spoken Language Processing (ICSLP '98) 1998 Vol. 6 Pages 2583-2586

  By Kazuhito Koishida, Gou Hirabayashi, Keiichi Tokuda, Takao Kobayashi
• Low Bit Rate Speech Coding Based on Mel-Generalized Cepstral Analysis

  Tokyo Institute of Technology 1998

  By Kazuhito Koishida
• Spectral Representation of Speech Based on Mel-Generalized Cepstral Coefficients and Its Properties

  IEICE Transactions on Information and Systems November, 1997 Vol. J80-A, No. 11 Pages 1999-2006

  By Kazuhito Koishida, Keiichi Tokuda, Takao Kobayashi, Satoshi Imai
• Spectral Quantization Using Statistics of Static and Dynamic Features

  1997 IEEE Workshop on Speech Coding for Telecommunications Proceedings. Back to Basics: Attacking Fundamental Problems in Speech Coding September, 1997 Pages 19-20

  By Kazuhito Koishida, Keiichi Tokuda, Takashi Masuko, Takao Kobayashi
• Efficient Encoding of Mel-Generalized Cepstrum for CELP Coders

  Proceedings, 1997 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) April, 1997 Vol. 2 Pages 1355-1358

  By Kazuhito Koishida, Keiichi Tokuda, Takao Kobayashi, S. Imai
• CELP Coding System Based on Mel-Generalized Cepstral Analysis

  Proceedings, 4th International Conference on Spoken Language Processing (ICSLP '96) October, 1996 Vol. 1 Pages 318-321

  By Kazuhito Koishida, Keiichi Tokuda, Takao Kobayashi, S. Imai
• CELP Coding Based on Mel-Cepstral Analysis

  Proceedings, 1995 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) May, 1995 Vol. 1 Pages 33-36

  By Kazuhito Koishida, Keiichi Tokuda, Takao Kobayashi, S. Imai