Benchmark Results of Up To 90% Hit Rate in 0 dB SNR Conditions at Less Than 150µW, a Fraction of the Power Consumption of Available DSPs
AONDevices, Inc., a private company focused on ultra-low power, high performance edge AI processors with integrated, application-specific inference algorithms, announced benchmark performance and power consumption data for its AONVoice™ edge AI technology.
In benchmarks, AONVoice™ processors deliver best-in-class recognition in real-world ambient noise conditions of up to 90% accuracy in 0 dB SNR conditions. Including front-end processing and all required memory, AONVoice™ processor consumes less than 150µW proven in 40nm ULP silicon. This power is measured in 100% constant speech conditions.
“Our technology evaluation of AONDevices’ low power voice and sound recognition IP was very positive. By adding artificial intelligence to our advanced wireless communication ICs, it offers the potential for voice activity and hotword detection at the edge in headsets, remote controls and other applications where performance, high noise-resilience and battery lifetime are crucial,” said Arend van der Weijden, Vice President Wireless, Audio and Voice BU at Dialog Semiconductor.
“We’re thrilled with our partnership with Dialog Semiconductor and proud to offer an end-to-end edge AI solution for integration in super low power wearables and consumer devices,” added Mouna Elkhatib, CEO of AONDevices.
The AONDevices solution provides complete edge AI design support to customers, from data acquisition to SOC level implementation. The tool suite includes dataset augmentation, training, and inference optimization to simulate chip level behavior. Each core can be configured for multiple custom wake words, voice commands, sound events and context detection. AONDevices has licensed its AONVoice™ edge AI processor technology to top tier semiconductor companies and is currently demonstrating best-in-class performance results in both ASIC and FPGA.