Acoustic Echo Cancellation is a standard yet tricky task: on a computer with laud speaker turned on, my microphone will capture both my voice and what played from the my computer laud speaker. You are asked to write a quality program that removes all from the laud speaker and only records my voice.
Your baseline delivery shall be a portable code runnable on both Windows and Android, given proper APIs for back-sampling speaker sound and for recording microphone.
What makes the code a bit more tricky is: actually we do NOT have the typical back-sampling of speaker. Instead, we have a loosely time aligned speaker out audio stream (from higher level Windows APIs). As a compensation, we are allowed to play out, together with the speaker audio, some ultrasound (at 20khz-22khz) which is identifiable from the recording mic (48khz sampling rate). You might want to take advantage of that by playing out some time sync ultrasonic tones.
We limit ourselves to a mono mic, for now.
Further details to be discussed in Q/A.
About the recuiterMember since May 20, 2018 Priya Parab
from Chieti, Italy