Far-end signal → speaker → mic picks up echo. AEC subtracts predicted echo.

What you're seeing

Speakerphones, conference rooms: audio from far end plays through speaker and gets picked up by the mic. Without AEC, far-end hears themselves with delay → painful echo.

AEC: estimate the speaker→mic impulse response (adaptive filter), subtract predicted echo from mic signal. NLMS, LMS, frequency-domain variants. Modern: deep learning.

★ KEY TAKEAWAY
Speakerphones need acoustic echo cancellation: predict the echo via adaptive filter, subtract it from the mic.
▶ WHAT TO TRY
  • Click Start — watch the three signals: far-end (blue), mic raw (red, with echo), mic clean (green).
  • Slide Echo strength up — without AEC, far end would hear a delayed copy of themselves.
← Back to Audio Engineering