Aspinity Inc. has launched its Voice-First Evaluation Kit (EVK2), a hardware/software development kit that demonstrates the company’s ultra-low-power edge processing solution for analog voice activity detection and wake word (preroll). The new dev kit shows designers of battery-operated devices how to move more of the audio processing capability to the edge without reducing battery life or system accuracy.
The EVK2 helps to ease integration of analog machine learning and analog data compression into battery-operated voice-enabled devices, such as hearables/wearables, smart speakers, and smart TV remotes, while enabling power savings without sacrificing wake-word detection accuracy in voice-enabled devices.
At the heart of the EVK2 is of Aspinity’s Reconfigurable Analog Modular Processor (RAMP) chip, or ultra-low-power analog machine learning (AnalogML) processor. The RAMP chip is a new architectural approach to system design that improves battery life in edge devices, which Aspinity says “determines what data is important enough to warrant high-processing earlier in the signal chain using minimal energy.” This is in comparison to alternative always-listening system architectures that digitize all sound data before wake word analysis.
The RAMP chip reduces power consumption by using near-zero power to analyze raw, unstructured analog microphone data at the start of the signal chain to determine if voice is present prior to triggering the wake word engine. This analyze-first approach minimizes the power-on time of the analog-to-digital converter (ADC) and wake word engine (WWE), increasing battery life by up to 10×, according to Aspinity.
The company also claims the RAMP chip is the first analog voice wake-up solution to continuously collect and compress (into ~2 kB of memory) the 500 ms of sound prior to the wake word (preroll) that is required by most WWEs in order to accurately determine that a command has been spoken.
In addition to the RAMP chip with voice activity detection and preroll collection, compression, and reconstruction algorithms, the EVK2 includes audio test files for quick start-up, a live audio testing option that uses an Infineon MEMS microphone, and integration with STMicroelectronics STM32H743ZI microcontroller that allows the testing of analog voice activity detection with or without preroll collection and delivery to a third-party WWE such as Amazon WWE. For more information about the technology and evaluation kit, here are links to a product brief (download) and video.
Advertisement
