voiceINTERconnect English Products Algorithms Echo Compensation

Echo Compensation

Phone applications, and in particular hands-free applications, are affected by acoustic feedback between the microphone and loudspeaker. This problem is usually caused by the so-called remote speaker. The voice signal of the remote speaker is released via the local loudspeaker, resulting in backcoupling into the local microphone.

Consequently, the remote speaker perceives his own voice with some time delay due to technical processing line. In case, the remote speaker is also using hands-free device, his loudspeaker signal is back coupling into the remote microphone, too. This scenario causes a continuous acoustic feedback loop, but in worst case, it is linked with unpleasant whistle. Simple devices try to solve the problem by using level scales. At the same time, only single speaker is transmitted. In case, both sides are speaking (cross-talk), stronger voice signal is preferred.

voice INTER connect is providing an advanced solution, which is enabling double talk but even so eliminating the undesirable echo signal. The Acoustic Echo Cancellation (AEC) solution of voice INTER connect is automatically detecting the characteristics of the signal processing path and is eliminating the echo part.

Basing on highly-adaptive configuration, the AEC solution is working under varying acoustic room characteristics (variable rooms with strong or without reverberation).

For further information go to:

Product features

upwards

AEC enables comfortable full-duplex communication albeit under difficult conditions:

Operation for room impulse responses (echo path delays) up to 512 ms delay
Optimized for car environment
Integrated Active Noise Reduction (NR) for noise suppression in disturbed environments (e. g. driving noise inside the car)
Low requirements concerning data memory and processing power - depending on length of room impulse response
Robust operation for cross-talk scenario without explicit double talk detector. An automatic adaptation to double talk is more reliable than an explicit detector and already given by the specific AEC algorithm
Reliable operation without temporary divergences under strong noise conditions

Technical features

upwards

ANSI C compliant
Application programming interface (API) allows easy implementation
Low consumption of data and program memory (e. g. ADSP 2153x at 64 ms room impulse response: approx. 40 MIPS, 5 kw program, 6 kw data)
Adjustable echo path: 64, 128, 256 or 512 ms
According to ITU-T G.167 recommendations
Negligible echo rest signal (ERLE: approx. 40 dB)
Continuous adaptation and fast convergence in real-time
Reliable operation even for small SNR (> 3 dB)
Efficient, integrated Active Noise Suppression (NR)
Direct interface for audio data (8/16 kHz, 16 bit). Both, sample-by-sample or block transmission are supported
Comfort noise