Acoustic Echo Cancellation

Hands-free Communication, Full-duplex, Intercom.

Connection between speaker signal and feedback to microphone

In a hands-free set, acoustic feedback exists between the loudspeaker and the microphone. This means that the microphone not only receives the speech signal of the near-end speaker , but also the signal emitted by the loudspeaker, which is the signal of the far-end speaker. This signal is called the echo signal. When the microphone signal is sent back to the far-end speaker, the far-end speaker will thus hear considerable acoustic feedback.

The far speaker hears his own voice time-delayed in his receiver. This delayed feedback impedes intelligibility and can be perceived very disturbing. If the far-away speaker uses a hands-free device too, his received signal through the loudspeaker feedbacks in the microphone as well and it results in a permanent feedback loop (audible as a whistle).

Conventional systems solve this problem with a loss control - but this allows only one speaker to speak at the same time. If both speakers speak at the same time (double talk), the stronger one will be preferred and the other one suppressed.

voice INTER connect offers an Automatic Echo Cancellation (AEC) System, which allows double talk and elimination of the echo at the same time. The AEC provides comfortable full-duplex communication even under difficult environmental conditions.

Fields of Application for Acoustic Echo Cancellation

In a handsfree set, acoustic feedback exists between the loudspeaker and the microphone. This means that the microphone not only receives the speech signal s(t) of the near-end speaker (the speaker in the car), but also the signal emitted by the loudspeaker x(t), which is the signal of the far-end speaker. This signal is called the echo signal y(t). When the microphone signal is sent back to the far-end speaker, the far-end speaker will thus hear considerable acoustic feedback.

Mode of Operation

Acoustic echo cancellation has the function of attenuating the echo signal y(t) within the microphone signal m(t). This is achieved using the signal of the loudspeaker x(t). But it is not enough to simply subtract the signal x(t) from the microphone signal m(t). The signal x(t) will be distorted by the loudspeaker, the microphone and the space in which the handsfree set is being used. The influence of the space or room is defined by the arrangement of sound reflecting surfaces within the car cabin. The acoustic echo cancellation algorithm tries to model these influences with a filter (echo cancellation filter). This filter estimates the echo signal y(t) and subtracts this from the microphone signal.

During this process, the echo cancellation filter is continuously updated by an adaptive algorithm. The outcome thereof is that the echo cancellation filter has on the one hand the ability to readjust itself to different loudspeakers, microphones and rooms, whilst on the other hand it also possesses the ability to track changes in the car cabin, for example if the near-end speaker moves. Effective attenuation of the loudspeaker signal wouldn't be possible without such an adaptive algorithm.

Limits of the Method

However, the echo cancellation algorithm isn't able to attenuate the echo signal completely. A certain amount of the echo signal always resides. The level of attenuation depends on the one hand on the adaptation algorithm used, and on the other hand on any external influences, such as disturbing signals like the speech signal of the near-end speaker s(t) as well as background noise like driving noise n(t). Furthermore, the quality of the loudspeaker also has an effect. The total harmonic distortion is particularly important. The smaller the total harmonic distortion, the higher the echo attenuation. To further increase echo attenuation, a postfilter is used, which further attenuates the residual echo. However, this has the disadvantage of distorting the speech signal s(t), whereas acoustic echo cancellation has no influence on the speech signal.