Classifying Objects Using Artificial Intelligence
In the landscape of today's automotive and aerial sensors, there is a huge blind spot that has severely restricted long range ability to detect, track, and classify objects for higher safety ratings. Cameras are low cost and high resolution but are unreliable because of increased processing to detect objects at long ranges and in bad weather conditions. At night or in bad weather their range quickly drops. Lidar is powerful but has so far failed to become low cost and low power enough for anything but highly specialized vehicles. It too cannot reach the ranges required for higher speed driving. Today's radar gets high marks for its low cost, reliability, decent range, and weather tolerance, but most self-driving car companies will tell you that camera and lidar consume the vast majority of their attention.
AWARE™ Integration with SPEKTRA™ for Autonomous Driving
AWARE™ integration with SPEKTRA™ for driver assist and autonomous driving: The reason for this is very simple: today's radar delivers vague, imprecise information about its environment. Metawave SPEKTRA™ radar solves this ‘lack-of-precision’ issue with its revolutionary beamsteering technology, and Metawave's unique fusion ML/AI platform based on lidars and cameras, AWARE™, uses the latest advances in deep learning to extract meaningful information from this one-of-a-kind resulting signal.
With highest real-time labeling accuracy, AWARE™ is paving the way to build radar brain that operates independently from other sensors.
SPEKTRA™ offers true analog RF beamforming combined with advanced MIMO signal processing, while reducing power, complexity, and cost. Its analog architecture is fast (µs speed), supports high resolution and high signal noise ratio (SNR), at least 10x more than digital beamforming (DBF), and significantly suppresses interence.
Machine Learning (ML) and Artificial Intelligence (AI) for Object Classification
By combining machine learning with SPEKTRA's unique sensing technology, we enable:
Object detection, classification, and tracking in real time using a proprietary deep neural network architecture
Resolution by training AWARE using our platform and AI algorithms to surpass conventional limits of precisoin
Adaptive Attention Network (AAN): Takes full advantage of ultra-fast beamsteering to tell the radar where to look next
Microdoppler signature extraction for enhanced pedestrian detection to perform object classification using the radar equivalent of one pixel over time