Reality AI
Reality AI is the leading product development environment for Edge AI using sensors. Reality AI offers AI-driven technology to help find the most cost-effective instrumentation, components, and placement - and to build, optimize and deploy Machine Learning models in your product’s firmware. Pre-built Edge AI sensing modules and Reality AI Tools® platform can be utilized to build own solutions.
Project details
In the Proof-of-Concept project, we evaluated the effectiveness of Reality AI’s algorithms to accurately detect, classify, and localize moving objects from their sound signatures in order to complement ADAS/AD perception systems. In addition, emergency vehicle sirens within 200 meters should be detected and classified with at least 95% accuracy.
Also, the algorithm should be able to localize vehicles within a 45-degree sector. The system consisted of Infineon's XENSIV™ MEMS microphones and AURIX™ MCUs, as well as Reality AI's software.
Common sensor suites used for ADAS/AD have “blind spots” in multiple scenarios that could be covered with sound-based sensing for automotive.
Utilized Reality AI’s machine learning algorithms to enable sound-based sensing for automotive exterior applications.
The system could successfully distinguish between two types of moving “road participants” (emergency vehicle and regular passenger car) and localized moving objects within a 45-degree precision accuracy.