Double talk detector for echo cancellation in a speech communication system

摘要

A speech communication system is provided that uses pitch information, pitch lags, pitch gains, energy and/or other speech characteristics about the outgoing speech and the unknown signal on a frame basis to determine if the unknown signal is an echo signal of the outgoing speech or if the unknown signal also contains speech from a second talker (double talk). Additionally, a plurality of frames of these characteristics of the outgoing speech signal and the unknown incoming signal may be buffered so that the analysis and comparison can be made more efficiently and quickly in the frame domain as opposed to a time domain. Multiple characteristics may be optionally weighted and then analyzed. The system and method may further determine a level of confidence, based on any criterion, in the determination that may then be used to adjust the gain of a filter.