AIStorm’s AIS2001 Monarch is a charge domain Waveform to Spectral Image converter primarily targeting key word spotting and sound spotting applications. It can take in waveforms from sensors such as a MEMs analog microphone input and produce a spectral image which can be processed by downstream microcontroller or AI engines.
Monarch's Revolutionary Capabilities
The AIS2001 Monarch is designed to bridge the gap between analog waveform generators and AI analysis. AI requires an image to process in application such as key word spotting (KWS), sound spotting (eg. glass break or gunshot), vibration analysis and certain biometric sensor applications. Rather than digitizing incoming analog information and performing costly DSP upon it as digital solutions would do, Monarch instead accepts and processes analog information directly using charge domain processing. The result is faster processing, lower silicon cost, and reduced power compared to ADC based digital solutions. Monarch accepts an analog waveform and outputs a digital image result through an SPI interface. To accomplish this the AIS339 includes: i) analog waveform input; ii) programmable filter coefficients over 32 bins; iii) SPI interface with programmable registers and image output; vi) <250uA charge domain processing of spectral information (can trigger an interrupt on amplitude and spectral components); v) 1.2v operation; vi) 32 pin 5×5 QFN package.
Key Features
Waveform to Spectral Image Converter (>10x Less than Digital)
Pair with AIStorm’s Sparrow AI Microphone for Best Results
Operation from 1.2V
Spectral Analysis Models Available
Compatible with Low Power Analog Microphones
Acoustic and Biometric Templates Available
SPI Interface
AIStorm Charge Domain Engines for Best Results. See SpectroMic
Compatible with Popular AI uC and GPU Engines
Produces AI Compatible Image
Multichannel Sensor and Stimulus Subsystem
High Order Filter
32 Bin Coefficient Programmable Waveform to Spectral Converter