hitachi-speech / EEND

End-to-End Neural Diarization

EEND (End-to-End Neural Diarization)

EEND (End-to-End Neural Diarization) is a neural-network-based speaker diarization method.

• BLSTM EEND (INTERSPEECH 2019)
  • https://www.isca-speech.org/archive/Interspeech_2019/abstracts/2899.html
• Self-attentive EEND (ASRU 2019)
  • https://ieeexplore.ieee.org/abstract/document/9003959/

The EEND extension for various number of speakers is also provided in this repository.

• Self-attentive EEND with encoder-decoder based attractors
  • https://arxiv.org/abs/2005.09921

Install tools

Requirements

• NVIDIA CUDA GPU
• CUDA Toolkit (8.0 <= version <= 10.1)

Install kaldi and python environment

cd tools
make

• This command builds kaldi at tools/kaldi
  • if you want to use pre-build kaldi

    cd tools
    make KALDI=<existing_kaldi_root>

    This option make a symlink at tools/kaldi
• This command extracts miniconda3 at tools/miniconda3, and creates conda envirionment named 'eend'
• Then, installs Chainer and cupy into 'eend' environment
  • use CUDA in /usr/local/cuda/
    • if you need to specify your CUDA path

      cd tools
      make CUDA_PATH=/your/path/to/cuda-8.0

      This command installs cupy-cudaXX according to your CUDA version. See https://docs-cupy.chainer.org/en/stable/install.html#install-cupy

Test recipe (mini_librispeech)

Configuration

• Modify egs/mini_librispeech/v1/cmd.sh according to your job schedular. If you use your local machine, use "run.pl". If you use Grid Engine, use "queue.pl" If you use SLURM, use "slurm.pl". For more information about cmd.sh see http://kaldi-asr.org/doc/queue.html.

Data preparation

cd egs/mini_librispeech/v1
./run_prepare_shared.sh

Run training, inference, and scoring

./run.sh

• If you use encoder-decoder based attractors [3], modify run.sh to use config/eda/{train,infer}.yaml
• See RESULT.md and compare with your result.

CALLHOME two-speaker experiment

Configuraition

• Modify egs/callhome/v1/cmd.sh according to your job schedular. If you use your local machine, use "run.pl". If you use Grid Engine, use "queue.pl" If you use SLURM, use "slurm.pl". For more information about cmd.sh see http://kaldi-asr.org/doc/queue.html.
• Modify egs/callhome/v1/run_prepare_shared.sh according to storage paths of your corpora.

Data preparation

cd egs/callhome/v1
./run_prepare_shared.sh
# If you want to conduct 1-4 speaker experiments, run below.
# You also have to set paths to your corpora properly.
./run_prepare_shared_eda.sh

Self-attention-based model using 2-speaker mixtures

./run.sh

BLSTM-based model using 2-speaker mixtures

local/run_blstm.sh

Self-attention-based model with EDA using 1-4-speaker mixtures

./run_eda.sh

References

[1] Yusuke Fujita, Naoyuki Kanda, Shota Horiguchi, Kenji Nagamatsu, Shinji Watanabe, " End-to-End Neural Speaker Diarization with Permutation-free Objectives," Proc. Interspeech, pp. 4300-4304, 2019

[2] Yusuke Fujita, Naoyuki Kanda, Shota Horiguchi, Yawen Xue, Kenji Nagamatsu, Shinji Watanabe, " End-to-End Neural Speaker Diarization with Self-attention," Proc. ASRU, pp. 296-303, 2019

[3] Shota Horiguchi, Yusuke Fujita, Shinji Watanabe, Yawen Xue, Kenji Nagamatsu, " End-to-End Speaker Diarization for an Unknown Number of Speakers with Encoder-Decoder Based Attractors," Proc. INTERSPEECH, 2020

Citation

@inproceedings{Fujita2019Interspeech,
 author={Yusuke Fujita and Naoyuki Kanda and Shota Horiguchi and Kenji Nagamatsu and Shinji Watanabe},
 title={{End-to-End Neural Speaker Diarization with Permutation-free Objectives}},
 booktitle={Interspeech},
 pages={4300--4304}
 year=2019
}