Hello? Acoustic Echo Cancellation & Voice Link DSP Workbench

Simulation Preset

Adaptive Filter (NLMS) Active

Enable AEC Engine

Filter Tap Length (N) 64 taps

Step Size Learning Rate (μ) 0.15

Double-Talk Freeze (DTD)

Acoustic Room Path

Acoustic Echo Coupling (ERL) 0.65x

Room Decay / Reverb (RT60) 280 ms

Near-End Background Noise -36 dB

VoIP Line Conditions

One-Way Transmission Latency 45 ms

Packet Jitter Dispersion ±12 ms

Simulated Packet Loss 1.5%

ERLE Echo Return Loss 24.2 dB

Residual Error ε 0.024

DTD Activity Far Only

Jitter Buffer Depth 60 ms

PESQ / Audio MOS 4.28 / 5

Real-Time Multi-Channel Signal Flow

Far-End x[n] ("Hello") Near-End Mic d[n] (Echo) Estimated Echo ŷ[n] Clean Output e[n]

Adaptive Filter Weights w[k] vs True RIR CONVERGED (98.4%)

Jitter Buffer Arrival & Depletion Spectrum LOSS: 0 PKTS

How "Hello?" Echo Cancellation Works

When you speak "Hello?" into a speakerphone, the sound travels through the far-end speaker, bounces off room walls (Room Impulse Response, RIR), and leaks directly into the near-end microphone. Without Acoustic Echo Cancellation (AEC), your own delayed voice bounces straight back to you.

The Normalized Least Mean Squares (NLMS) filter continuously models the room acoustic path h[n] by calculating weight vector updates: w(n+1) = w(n) + μ · e(n) · x(n) / (||x(n)||^2 + δ). If the near-end speaker also speaks at the exact same moment (Double-Talk), the Double-Talk Detector (DTD) temporarily freezes weight updates to prevent the filter from diverging.