Alango's speech dereverberation technology detects and attenuates reverberant speech within an audio signal while preserving the direct speech to make it more intelligible.
When speech is acquired inside an enclosed space, such as conference room or car cabin, the resulting signal contains the sum of all reflected sound signals, which are delayed and attenuated along their indirect paths to the microphone; this effect is known as reverberation. In voice communication systems, untreated reverberation in a speech signal can reduce intelligibility and strain the ear of the listener.
Illustration of speech acquired by a speakerphone inside a reverberant room with ambient noise
An example of this is the speech signal acquired by a table-top conference speakerphone within a resonant conference room, where multiple near-side users might speak at considerable distance from the microphone. Ambient noise can also reverberate within the room, compounding the problem even further.
Detecting reverberation in the speech signal
When speech is acquired in a reverberant room, the speech dynamics are changed; speech decay patterns are lengthened. To detect reverberation, the algorithm continuously estimates the speech decay length. If the decay tail exceeds a preset amount, dereverberation is activated.
Use control buttons on the interactive screenshots to play/stop/navigate
Raw signal, reverberant speech recorded in conference room:
Processed signal, after noise suppression and dereverberation
- Conferencing mics and speakerphones
- Bluetooth headsets
- Mobile phones
- Automotive Handsfree
- Hearing Aids
- Voice-controlled devices
Alango’s dereverberation technology is low latency and computationally efficient, making it suitable for real-time voice communication enhancement on devices such speakerphones, headphones, and headsets. The technology is integrated within the Noise Reduction processing block in VCP (Voice Communication Package) and eVCP (Extended Voice Communication Package).