Abstract:
An acoustic echo cancellation (AEC) method developed for voice over IP (VoIP) hands-free applications is proposed. This method can effectively reduce undesired acoustic echo arriving at a microphone from a loudspeaker and emphasize the target talker's voice when near-end and far-end talkers speak simultaneously (i.e. double-talk), irrespective of smartphone/tablet device models. This method mainly involves cancellation of non-linear acoustic echo caused by loudspeaker distortion, residual echo reduction robust against echo-path change, and estimation of pure delay resulting from both room echo and audio input/output buffers. The experimental results show that the proposed AEC method reduced more than 40 dB of undesired echo for every smartphone or tablet used for the evaluation. This indicates that the performance of the proposed AEC method does not depend on the difference in the acoustic characteristics of individual devices.
