Asunny2019 / 2023_AI_Academy_ASR

2023 AI Academy ASR

딥러닝을 이용한 음성인식 기초 실습

0. Tutorial

Tutorial for python and data science packages

• python review
• numpy
• matplotlib

1. Audio file handling

Audio file handling using torchaudio

• Load audio file(torchaudio.load)
• Feature extraction(Mel-spectrogram, MFCC)

3. Audio Classification using MLP

Audio MNIST classification using MLP(torch.Linear)

4. CTC

Simple Exercise(model training using CTC loss) for Connectionist Temporal Classification

5. Whisper

Exercise using OpenAI - Whisper and Gradio

6. E2E ASR model finetune with Nemo

Quartznet Model finetune with Nemo(English to Korean)

7. WFST

Exercise for WFST using k2

• C,L,G transducer
• composition, determinization

Libraries

• PyTorch : pytorch/pytorch
• NeMo : Nvidia/NeMo
• TorchAudio : pytorch/audio
• NumPy : numpy/numpy
• matplotlib :matplotlib/matplotlib
• Whisper : openai/whisper
• gradio : gradio-app/gradio