Speech denoising
Background noise drowns out the voice.
Your product has to carry a voice over traffic, chatter and machines. We do it with AI speech denoising, and help you build it in.
Last updated
Request a demo
How it works
From a noisy microphone to a clear voice
Noise changes from second to second and often sounds like speech. A neural network estimates which parts of the sound belong to the voice and removes the rest, as the person speaks.
Your system: Sound
A voice, with traffic, chatter or machines around it.
Your system: Microphone
The microphone in your product.
Our solution: Spectrum
Splits the sound into slices of time and frequency.
Our solution: Speech denoising
Keeps the voice and removes the noise in each slice.
Your system: Call
Carries only the voice, without the noise.
What it handles
Voices, not background noise
We tune it to the noise your product meets, then check each recording to make sure the voice is kept and the noise is gone.
What it keeps
- Soft speech
- Quiet voices close to the level of the noise.
- Different voices
- High and low, fast and slow.
- Every word
- The start and end of each word, not just the loud parts.
- A step away
- Speech from close by or a step back.
- Your microphone
- The microphone and enclosure in your product.
What it removes
- Traffic
- Engines, horns and road noise.
- Chatter
- Background voices in cafés and offices.
- Machines
- Fans, motors and grinders.
- Clatter
- Cups, cutlery and doors.
- Hum
- Air conditioning and steady electrical hum.
Applications
Where it's built in
Your product gets a clear voice, with the noise taken out.
Headsets
Clear calls from busy offices, streets and shop floors.
Intercoms
Door and gate entry beside loud traffic.
Conference systems
Meeting rooms with fans, typing and chatter.
Evaluation
Interested in this solution?
Get in touch with our engineering team to evaluate it for your system.
Evaluation platform. Firmware and software are under development; each use case requires application-specific validation.
Other use cases
More on denoising
See how we pick out spoken commands.
Keyword spotting
Cleans the voice first, so spoken commands are recognised over the noise around them.
