search

fpga-open-speech-tools / simulink_models

Simulink models for speech and audio processing
MIT License
4 stars 4 forks source link

Noise Suppression Properly Verify #70

Closed dack-fe closed 3 years ago

dack-fe commented 3 years ago

The Noise Suppression model contains additional scripts - https://github.com/fpga-open-speech-tools/simulink_models/tree/dev/models/noise_suppression/MATLAB_scripts

What are these scripts? Can they be removed?

tvannoy commented 3 years ago

They were during development and testing. Essentially, the algorithm was implemented in MATLAB first, then implemented in Simulink.

They can be removed. They could perhaps be incorporated into verifySim; the output of adaptiveWienerFilt.m should be close to the output of the Simulink model, but I don't believe it was ever rigorously verified.

fe-wickham commented 3 years ago

I think it would be worth it to get the scripts used in verifySim as ideally we can show in our example projects actual verification going on. If any of the scripts end up not being needed for verification we can get rid of them then

tvannoy commented 3 years ago

Agreed. This would be a good example of a non-trivial verification, versus something trivial like the simple gain. adaptiveWienerFilt should be the only file needed for verification.

We can either keep adaptiveWienerFilt as a separate file or put the functions into verifySim as local functions. Do we have a preference here?

fe-wickham commented 3 years ago

Personally I am preferential to keeping something as complicated as adaptiveWienerFilt as a separate file just to keep the verifySim file still easy and quick to parse for readers rather than getting bogged down in the details if they aren't interested.

Its also a bit better practice as it enables to us to do more complex validation when it is used as part of a bigger model, ie the ANM components and validating the larger and larger sections of it.

I have updated this issue's title to reflect that the verification should be updated