Wake word engine available for CEVA DSPs
CEVA and Retune DSP have announced that the Retune VoiceSpot Wake Word Engine is now optimised and available for CEVA’s audio/voice DSPs, including the CEVA-BX, CEVA-X, and CEVA-TeakLite families. CEVA and Retune DSP are already engaged with a lead customer, deploying this collaborative solution in an upcoming smart home device.
VoiceSpot is a high-performance embedded wake word and small vocabulary ASR engine, enabling voice-wake and command and control on memory-constrained, battery-operated edge devices. VoiceSpot achieves high performance levels with very small 30 to 35KB typical model size (depending upon the specific wake-word), with a total memory footprint of under 80KB.
CEVA’s audio/voice DSPs are widely deployed in battery-operated edge devices for voice activation and control, including TWS earbuds, smartwatches, action cameras and smart home appliances. These DSPs are suited for hosting wake word engines like VoiceSpot at extremely low power standby, while also being able to run complex voice front-end algorithms such as echo cancellation and beamforming to improve voice intelligibility and pickup quality.
“CEVA’s wide-reaching success in powering intelligent edge devices with their industry-leading audio/voice DSPs makes them an ideal partner for our VoiceSpot wake word engine,” said Ulrik Kjems, CEO of Retune DSP. “Our small memory footprint solution delivers outstanding voice-wake performance when implemented on the CEVA DSPs. Together, we can accelerate the adoption of voice command and control in smart devices, no matter the form factor or power constraints.”
“Retune DSP is a valued algorithm partner to CEVA, and we’re pleased to announce our collaboration with them around their VoiceSpot technology,” added Moshe Sheier, Vice President of Marketing at CEVA. “Voice is gaining rapid momentum as a must-have user interface for many smart devices and VoiceSpot running on our audio/voice DSPs provides a compelling offering for this burgeoning market.”
CEVA's scalable audio and sensor hub DSPs are optimised for sound processing applications ranging from always-on voice control up to multiple sensors fusion. They have been specifically designed to tackle multi-microphone speech processing use-cases, high quality audio playback and post-processing, and on-device sound neural network implementations.
In addition, a large 3rd party ecosystem of audio/voice software, hardware and development tools companies have optimised their solutions for CEVA DSPs, for a wide array of use cases and applications.