Add support for Real-time Noise cancellation in all DeepSpeech Inference Examples (Feature Request)

[Sadam1195 (Sadam Hussain Memon)]

Hi all, I have few ideas to clean audio in real-time and make it noise free. In fact I have tested code by recording and playing back back audio with noise canceled. I was wondering if anyone here would be ineterested to assist me add my code in DeepSpeech-examples repo for

mic_vad_streaming.py

example. As we say garbage in garbage out so better/clean input will certainly make quite a lot difference to the output of the model.

Best,

[dkreutz (Dominik)]

You may want to look into following discussion about the same/similar topic:

The noise reduction componet in the pre-transcription processing

Hello, I’m was thinking about to implement noise reduction component before the actual transcription stage. I’m curious if this will improve the accuracy, since the audio will be cleaner. Has anyone already did it? Did you implement something like a low-pass filter or something similar? And how this impact the transcription? Is this something like a good practise in Speech Recognition or rather not? Thank you in advance!

Need help with audio cleaning/de-noising

Hello everyone, I need help/ideas about what I can use for de-noising audios so I can improve the model accuracy. I have already tried RNNoise, model output for some audios gets better and for some it gets worse(which is not helping). So, can you guys suggest anything else that I can do? It’ll be really helpful. Following are the details I’m working with: Acoustic Model : 0.7.1 released by Mozilla Scorer : Custom Data : Conversational data. Customer support. I am splitting the audio call i…

[Sadam1195 (Sadam Hussain Memon)]

Yes, I have read those threads but I was thinking maybe they need test out those and integrate their code with DeepSpeech-examples or particularly with streaming examples to get better output. Hence the “feature request”

[lissyx ((slow to reply) [NOT PROVIDING SUPPORT])]

@Sadam1195

Hence the “feature request”

Please understand the examples are maintained by people who are writing them. So if you want the feature, either contact it’s author or write it.

Please also note that the examples are mostly here to help people see how to use the API in various situations.

[Sadam1195 (Sadam Hussain Memon)]

Hmm, understood. I was under the impression that both were managed by Mozilla and probably same team of devs. Thanks for clearing that up. I will write that up myself and maybe send them a PR.