FADE demo

This is a demo to re-generate the training process of FADE

请注意，本项目核心思路来源于FADE识别SRTS模型，本项目近乎为伪代码，需要后续完善

pip install -r requirements.txt

run python data_generation.py to generate speech and noises and mix them in different SNR

run train.py to train the whisper model in different SNR

run test.py to test the recognition score and get the RRM matrix

run recognition_map.py to calculate the SRTS (under hesitation)

run correction.py to modify the speech recognition function to the average SRT point