Parametric Cnn for Radar Processing

Simple SummaryContent extracted from patent full text and abstract with AI.

This patent discloses a method and system that uses a parametric, constrained L-dimensional convolutional neural network (CNN, L ≥ 2) to process raw digitized signals generated by a millimeter-wave radar. Unlike traditional radar signal analysis, which typically involves extensive signal preprocessing and feature extraction, this approach enables direct processing of raw radar data by the neural network. Specially designed CNN filters (such as 2D sinc or Morlet wavelets with learnable parameters) allow efficient and accurate extraction of information, such as object classification and location, from radar signals in real time.

Use CasesContent extracted from patent full text and abstract with AI.

• Human activity recognition in smart homes and eldercare (e.g., fall detection, movement monitoring)
• Gesture sensing for touchless human-machine interfaces
• Automotive radar for detecting and classifying pedestrians or other vehicles
• Physical security and surveillance (detecting presence, counting people)
• Industrial automation (monitoring machinery movement or presence of workers)
• Robotics and drones (navigation and environment sensing)
• People counting in public spaces or offices
• Non-intrusive monitoring in sensitive environments requiring privacy (hospital rooms, bedrooms, etc.)

BenefitsContent extracted from patent full text and abstract with AI.

• Eliminates need for conventional radar signal preprocessing, reducing computational complexity and system latency
• Enables effective and direct end-to-end learning from raw radar data for increased classification accuracy
• Supports real-time processing for rapid response applications
• Preserves privacy compared to camera-based systems as it does not capture visual information
• Achieves faster neural network training and model convergence by using constrained, parametric filter shapes
• Reduces model size and resource requirements, making hardware implementation practical in embedded systems
• Highly flexible—can be adapted to classify a wide range of targets and activities beyond human monitoring

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

In an embodiment, a method includes: transmitting a plurality of radar signals using a millimeter-wave radar sensor towards a target; receiving a plurality of reflected radar signals that correspond to the plurality of transmitted radar signals using the millimeter-wave radar; mixing a replica of the plurality of transmitted radar signals with the plurality of received reflected radar signals to generate an intermediate frequency signal; generating raw digital data based on the intermediate frequency signal using an analog-to-digital converter; processing the raw digital data using a constrained L dimensional convolutional layer of a neural network to generate intermediate digital data, where L is a positive integer greater than or equal to 2, and where the neural network includes a plurality of additional layers; and processing the intermediate digital data using the plurality of additional layers to generate information about the target.