Double-Talk Detection for Digital Network Echo Cancellers
in Presence of Signal Clipping
Awais M Hussain
Abstract:
The standard Geigel double-talk detector used in digital network echo
cancellers begins to fail in presence of signals produced by high-power
speech sources. A signal with high amplitude gets clipped at the
A/D converters. Filtered through the analog hybrid, the clipped signal
regrows some of its peaks resulting in false double-talk detection. The
Geigel algorithm is augmented to work under such conditions.