# Signal chain for voice calls including adaptive noise cancellation, adaptive echo cancellation, and automatic gain control and their algorithms

I am trying to understand the entire signal chain and all the algorithms associated with adaptive filtering as mentioned in the case above. From my understanding:

• Adaptive noise filtering (ANF)- can be performed with the help of Weiner filters, wavelet packet based auditory masking or subspace projections -- all these help in removing correlated noise. For uncorrelated noise, we can use a low pass filer? Gaussian filter bank in some way (need some guidance on this)?

• Adaptive Echo cancellation (AEC) - draws some speech from the far end speaker and modulates it so it sounds like the "feedback" from the near user so it can be successfully cancelled out from the near-end input (uses NLMS and RMS)?

• Automatic Gain control (AGC)- Look-ahead limiter?

From my understanding, the chain goes as follows:

Near end $\rightarrow$ Uncorrelated noise removal $\rightarrow$ Correlated noise removal + AEC $\rightarrow$ AGC.

Also how do Voice Activity Detection (VAD) and Discountinuous Transmission (DTX) work in this chain?

Is this correct? Can someone explain in detail about this?

• Questions on this site are supposed to be specific; try breaking your question up into several questions. Also, try to avoid asking for material that can be readily found on textbooks or sites such as wikipedia: look into the material first, and then ask about what you don't understand. – MBaz Aug 19 '17 at 23:22
• I've read about these concepts online and several papers published too. The questions are Audio DSP specific and overlap with the Telecomm category, so I don't see why I cannot post it here. Also the signal chain is confusing on all online sites and not streamlined, that is why I posted the question on here to get some direction. Would you happen to have the answer? – Akhilesh Rao Aug 19 '17 at 23:52