From the main acoustic impairments present in full-duplex audio communication systems, acoustic echo is the one that contributes to quality degradation the most. Acoustic echo occurs when the sound produced from one or multiple loudspeakers ➊ is captured by the microphone ➌ of the very system, and is sent back to the far-end of origin. The goal of acoustic echo reduction is two-fold: to reduce the echo components ➊ in the microphone signal ➌ to the extent that they become inaudible, and to preserve the main near-end sounds ➋ captured by the microphone clean and clear. Even after decades of research and development in this problem, the reduction of acoustic echo has not matured at the pace than other audio technologies have.
In a virtual conference, both loudspeaker and microphone should be active all the time, allowing all parties to talk and listen at any time. However, in order for that high-quality audio conversation to be possible, the videoconferencing system must remove efficiently the loudspeaker audio that is captured by the microphone, hence transmitting the voice of the people in the room only.
Check below how our solution performs in this difficult acoustic environment, plagued with multiple long-lasting echoes, under a “stress test” with permanent double talk.
Experiments conducted obtained in a similar deployment and room as that shown in the picture. Use Google Chrome for better experience.
In hands-free phone calls inside the car, the partner’s voice played by the loudspeaker is picked up by the very microphone. In order to accomplish a high-quality conversation, the car audio system must remove efficiently that echo, so that only the voice of the driver (or passengers) go through to the other end. Our solution performs exceptionally in this changing acoustic environment.
In order to detect voice commands from the desired sound source (green), the audio front-end of a voice-commanded device must remove the system’s audio (yellow) as well as a nearby sound interference (red), all captured by the microphone (click 1). Our technology removes the system music efficiently (click 2), and the sound interference (red) is further cancelled (click 3) with the integration of our technology and beamforming in an array (blue) with four-microphone . The “Alexa” keywords become undistorted and free of music and external interferences (compare 1 versus 3).
Experiment conducted in the scenario shown in the picture. Use Google Chrome for better experience.
In active noise cancellation for earphones, the noise (click 1) is picked up by the external microphone (click 3), so that the music (click 2) plus the “anti-noise” are simultaneously played by the loudspeaker (click 4). This digital processing is driven by the signal picked up by the internal microphone (click 5) resulting in the noise within the ear canal being acoustically cancelled while only the music arrives the eardrum. Our technology is capable of estimating the primary path (external microphone to internal microphone), the secondary path (loudspeaker to internal microphone) and the feedback path (loudspeaker to external microphone) simultaneously and altogether, which guarantees perfect noise cancellation for any user regardless of the earphones usage and user’s ear shape.