scalable multiple tts识别

plan1:

  1. collect single audio files from multiple tts service

  2. concat single audio file into 8 characters audio files

  3. convert audio to spectrogram using sox command

  4. use train.py to train the model

plan2:

  1. collect single audio files from multiple tts service

  2. train model for single character

  3. split test audio with 8 characters into single character file based on silence

  4. use classify.py to recognize single character

  5. combine the results from single character