ggueret / noisekit Goto Github PK

• Be able to connect many input and output over the network.
• Add swarm(s) configuration and mapping policies.

By default, the mitigator must handle frequencies like an human ear (20Hz - 20kHz).

• write noisy samples to audio file(s)
• exposing environment metrics to graphite to have an history

I've used Click for many project and he always brought me much more satisfaction than argparse.
Click allows to a better code structure, he is lightweight, popular, actively maintained and used in big OSS projects.

Implementing Click is a good opportunity to review the entire codebase and to refactor, which is perfectly in line with targets of the v1.

Provide configuration for humans.

Currently, the consumer will read the input from PyAudio as chunks. Every of theses chunks are analysed to find the RMS amplitude and the major frequency.

Then, an action is taken, to respond (or not) with a specific sound regarding the amplitude.
So, this behavior do not actually analyse the noisy sample from start to end, it just reflect a period of it.

If a noisy sample is longer then the chunk size, and start with a low level of amplitude but became higher on the next chunked sample, the mitigator will response for low too precocious when it may be higher to really reflect the nature of noise.