xmos/audio_test_tools

Implement DSP component performance framework

keithm-xmos opened this issue · 2 comments

  • Implement AEC performance metrics
  • Write script to process a dataset json file & output a report
  • Finish annotating the mono & stereo AEC training audio files
  • Generate baseline report for the mono & stereo AEC training dataset
  • Record, annotate & baseline mono & stereo AEC validation audio files
  • Document the dataset json format and AEC performance metrics

ASJ has requested additional training data:

  • music playing, lots of people constantly walking around the whole time, silence + alexas
  • no music playing, silence + alexas
  • very quiet music playing, i.e. mics can hear it, voice is easy, silence + alexas, echo path change after 30 seconds
  • very loud music playing, silence + alexas, echo path change after 30 seconds

Which makes 1+2+2+2 additional recordings.

Need some recordings with noise that is not alexa utterances.