VoiceSeeker far-field voice UI
VoiceSeeker is a DSP library designed for creating voice assistant type applications that use voice triggers and far field speech recognition as well as conventional hands-free communications.
- Easily design systems that appear to be "always on", listening for voice commands from a user
- Voice trigger from far field (> 5 meters)
- Voice trigger in noisy conditions
- Voice trigger during stereo music playback
- Support for up to 11 microphones (depending on available MIPS and memory)
- Contains hands-free communication mode, with multi-mic 360 degree target tracking beamforming and noise reduction combined with echo cancellation and non-linear echo residual removal algorithms
The aim of the library is to enable customers to easily design high performance voice trigger and voice recognition systems, augmented with hands-free communication mode.
Due to our unique multi-microphone beamforming technique, we have managed to make the 4-microphone beamforming voice trigger will even outperform the Amazon Echo's 7 microphone setup.
The library is even flexible enough to allow custom microphone configurations. However, microphone array design changes can have a large effect on the resulting system performance level, we always recommend consulting with us about any changes to our set of recommended geometries.
As a guideline, we have analyzed some common examples of arrays and their predicted beamforming performance in a diffuse noise field. This gives an indication of performance obtainable using different arrays. Follow this link (opens new window). As our beamforming is fully adaptive, these plots are meant to give an indication of attainable performance levels, they do not represent the actual performance in a given setup, which may be better or worse depending on the exact acoustical environment.