SBIR Phase II: Low power hardware-software subsystem for intelligent sensory stream analysis

The broader impact/commercial potential of this project is to significantly reduce the development effort for continuous sensing applications and to enable deploying them in a wide range of battery-constrained sensing devices, including personal mobile devices like smartphones, tablets, and smart watches; wearable health monitoring devices for tracking EKG/EEG and other vital signals; and remote sensing devices for monitoring structural integrity, emission levels, pollutant concentrations, or seismic data. In these applications, continuous sensing triggers context- and location-aware computation or communication in response to sensory stimuli in the device?s environment. Without dramatic innovations in the development of ultra-low power sensory processing, continuous sensing will remain a niche application limited to environments with a stable and plentiful power source. In contrast, this project will demonstrate the viability and potential widespread deployment of continuous sensing in mobile and remote environments. The ability to flexibly deploy continuous sensing for these and other applications has the potential to revolutionize these markets and create entirely new and unforeseen application domains, dramatically altering the extent to which cyber-physical systems can interact with their physical environments. This Small Business Innovation Research (SBIR) Phase 2 project will develop a complete software and hardware platform for developing and deploying applications that identify and respond to spatial and temporal trigger signatures (events of interests) in continuous sensory data streams. Current sensory processing platforms rely on a power-hungry application processor to analyze continuous sensory streams of real world phenomena to identify trigger signatures, and application development is cumbersome and requires significant domain expertise. The proposed approach offloads continuous sensing and trigger signature detection from the power-hungry application processor to a recognition engine deployed either in firmware on a low-power microcontroller or as a custom hardware accelerator. Furthermore, an integrated software development toolkit will be provided to streamline implementation of sophisticated sensing applications, leading to widespread adoption and future availability of always-on, sensor-based applications. The result of this project will be a software implementation of the novel neurally-inspired algorithm/firmware optimized to be deployed on currently available smartphones, a software toolkit that enables easy and rapid development of always-on sensing applications, and a novel ultra low power, neurally-inspired sensory data preprocessor.