In far-field hands-free communication systems, non-linear acoustic echoes can often be a problem. This is due to the need to strongly amplify the loudspeaker and microphones signals in order for the user of the system to hear and to be heard. Due to the close proximity of the microphone to the loudspeaker, acoustic echoes from the loudspeaker can easily saturate the microphone signal. Saturated echoes cannot be modeled by any linear acoustic echo cancellation algorithm. In addition, they make double-talk decisions impossible so even the non-linear processor (NLP) has a hard time deciding whether it needs to attenuate the signal. Therefore, the hardware gains of the loudspeaker and microphone need to be modified to prevent saturation. It is recommended to lower the microphone gain first because microphone signal can be gained back up after the acoustic echo canceller (AEC) with an automatic gain control (AGC).
Once the hardware gains have been tuned to avoid saturation, the system must be evaluated for any other non-linear echo sources. When the loudspeaker and microphone are enclosed in the same device, the mechanical vibrations of the loudspeaker can be picked up by the microphone, especially when the loudspeaker is driven hard. Performing various sound tests help identify the strongly coupled frequencies in the loudspeaker microphone enclosure response that are responsible for non-linear echoes. To help prevent these non-linear echoes, the loudspeaker signal can be filtered to attenuate those troublesome frequencies. In other words, the loudspeaker signal is custom pre-conditioned for the far-field hands-free communication device to achieve the best AEC and system performance. The figure below shows a block diagram for a pre-conditioned audio system with AEC and AGC.
VOCAL Technologies, Ltd., along with their AEC offerings, provide testing and tuning services to optimize it’s performance. Please contact us today to learn more.