Our low latency audio source separation technology removes the interfering noises that make effective speech recognition impossible. In tests, we improved the sentence recognition rate for a leading home assistant platform from 22% to 91% in noisy conditions. But it’s not just the performance that is remarkable; the technology has the following key features necessary for any practical product:
The array geometry is not critical - we can work with 3 to 8 microphones with any spacing from 2cm to 30cm apart.
There is no calibration or algorithm training requirement, minimising your production costs.
The algorithm has a low latency (<10ms) mode for time critical applications.
The algorithm automatically compensates for microphone occlusion and gain mismatch.
The algorithm gives consistent, natural sounding performance across a wide variety of real world conditions.
The highly efficient algorithm and coding is designed to obtain the best possible performance from the available processing power.