In full-duplex voice communications,
the far-end speaker hears his own voice
with some delay depending on the environments,
vocoder delay and impedance mismatching.
The echo phenomenon is cause by the coupling
between loud speaker and microphone(acoustic
echo) and electrical leakage of current
in the hybrid(line echo).
Adaptive filtering is one
of the most common solutions to the problems
of line and acoustic echo. The block updated
NLMS algorithm offers a good tradeoff between
computational load and performance.
General Feature
ITU G.165 Performance Test Specification
(Line Echo canceling)
ETSI Specification (Acoustic Echo Canceling)
Transversal Adaptive FIR Filter with
NLMS block update algorithm
Double Talk Detection to avoid divergence
of the adaptive filter
Active channel Detection
16ms/32ms/64ms/128ms Echo Path
250ms Convergence Convergence Time
Test
Description
ITU
G.165 Requirements
Infinior
performance
Final echo return
loss after convergence : Single-talk
mode
Steady-state residual
echo level after convergence without
near-end signal
-40dbm0
< -40dbm0
Convergence rate
:Single-talk mode
Combined echo loss
after 500ms from initialization
> 27 dB
> 38 dB
Leak rate
Degradation of residual
echo after 2 minutes from time all signals
are removed from converged EC
< 10 dB
0 dB
Infinite return loss
convergence
Returned echo level
500ms after echo path is interrupted
,TCP/IP
stack...
