I am working on Noise suppression and cancellation in VoIP media engine. I am looking for noise cancellation & suppression algorithms and its implementation available with real-time low latency algorithms.

I had worked with webRTC noise cancellation implementation and it works well, but I want to explore more on this. I googled this and am not getting any concrete output. I need help from the community and experts to get better an idea of available noise cancellation and suppression algorithms. Below are findings:

Noise cancellation available implementation:

  • WebRTC
  • Speex

Noise cancellation available Algorithm:

  • Least Mean Square Algorithm

finding more on Github is giving lots of results, but needs advise from experts to the way to proceed for choosing Noise cancellation & suppression algorithm for VoIP with low latency.


2 Answers 2


You could use DSP algorithms based on deep learning techniques like this scenario:

You have clean voice Sig.1 .

You have known different types of noise Sig.2 .

Merge Sig.1 and Sig.2 then Sig.3 generated (noisy speech).

Feed it Sig.1 to deep learning based software also Sig.3 then make training the system.

Then obtained coefficients should be close enough to Sig.2

In this way you will have a system that could predict the noise type in real time and generated the anti noise signal which it opposite in phase (180) and add the microphone signal before send .


If you are looking for free open-source ones, the above is the only available list (IMO). Otherwise, there are more advanced options. For example, check out SoliCall's noise reduction software. They offer both PNR (profile-based noise reduction) and RNR (reference-based noise reduction).

  • $\begingroup$ Looking for Open-source only. so i can check performance and integrate it $\endgroup$ Commented May 3, 2018 at 10:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.