Method and apparatus for testing echo canceller performance

Abstract

A test system measures performance of telephone network echo cancellers using a primary criterion of estimated user annoyance due to audible returned echo. The invention generates live telephone calls, uses real speech samples as stimulus signals and provides tail-circuit emulation using actual measured telephone tail-circuit impulse responses. These features provide better 'real-life' test conditions for the echo canceller system under test than current ITU standard test methods. Two methods are employed for echo canceller performance evaluation via metrics of estimated user annoyance due to echo. Energy-based method employs point-by-point comparison of talker speech and talker echo signal energy envelopes and uses variable energy thresholds for estimation of echo audibility. A perceptual-model based method uses a Perceptual Speech Distortion Metric (PSDM), such as ITU P.861, in an unique configuration to estimate user annoyance due to audible echo. Echo canceller performance is tested under both single-talk and double-talk conditions. Innovative application of the PSDM method in double-talk tests allow estimation of quality of received double-talk speech.