CEVA and Retune Unveil VoiceSpot Wakeword Engine
Delivers High-Performance Voice Wake, Command, and Control Functions
for Memory-Constrained Battery-Powered Edge Devices


On the 28th, CEVA announced the launch of the Retune VoiceSpot WakeWord Engine optimized for its audio/voice DSPs including △CEVA-BX, △CEVA-X, and △CEVA-TeakLite through collaboration with Retune DSP.

VoiceSpot is a high-performance embedded automatic speech recognition (ASR) engine that operates on wakewords and a small number of words, providing voice wake, command, and control functions in memory-constrained battery-powered edge devices.

While varying by wakeword, typical model size is 30–35KB, with total memory footprint under 80KB. CEVA's audio/voice DSP is well-suited for battery-powered edge devices including wireless earbuds, smartwatches, action cameras, and smart appliances that can be activated and controlled by voice.

Additionally, it is optimized for hosting wakeword engines such as VoiceSpot in low-power standby states, and executes complex voice front-end algorithms like echo cancellation and beamforming to improve speech clarity and recognition quality.

Meanwhile, CEVA's scalable audio and sensor hub DSP is useful for acoustic processing applications ranging from always-on voice control to multi-sensor fusion, and is specifically designed to handle multi-microphone speech processing use cases, high-quality audio playback and processing, and on-device acoustic neural network implementation.