People Counting Based on Radar Measurement and Data Processing in a Neural Network

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

This invention describes a computer-implemented method for accurately counting people in a scene using radar sensors and advanced neural network data processing. Instead of relying on cameras (which can be affected by lighting and privacy concerns), it processes radar data to create detailed range-Doppler measurement maps (representing both large and subtle human movements) and feeds these into a specialized neural network. The network uses separate processing pipelines for macro-Doppler (large movement) and micro-Doppler (small movement) features, combines their results, and outputs a reliable estimate of the number of people present—even when people are close to each other or stationary.

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

• Monitoring the number of people entering or exiting at doors, corridors, or building entrances.
• Foot traffic analysis and customer flow management for retail stores, shopping centers, museums, and entertainment venues.
• Occupancy-based building automation (e.g., adjusting heating, ventilation, or air conditioning based on people count).
• Public safety management, such as enforcing crowd limits during events or in response to health regulations (e.g., during pandemics).
• Smart transportation hubs for tracking occupancy in buses, trains, or stations without privacy invasion.
• Security and intrusion detection in restricted or sensitive areas where cameras may not be suitable.

BenefitsContent extracted from patent full text and abstract with AI.

• Accurate people counting in varied environments, including low-light or obstructed scenarios where cameras are less effective.
• Enhanced privacy, as radar does not capture identifiable images of individuals.
• Ability to distinguish and count multiple people, even if they are close to each other or moving minimally.
• Robustness to different sensor poses and scene geometries, supporting flexible deployment.
• Improved reliability using advanced neural network architectures tailored specifically for radar features.
• Supports real-time use with smoothing and tracking filters (e.g., Kalman filters) for continuous monitoring.
• Facilitates energy savings and operational efficiency in smart buildings through reliable occupancy detection.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

Various examples relate to people counting based on radar measurements. Computer-implemented example methods comprise: based on a radar measurement dataset obtained by a radar measurement of a scene, determining at least one measurement map indicative of features of one or more persons included in the scene; and processing the at least one measurement map in a machine learning or neural network algorithm, the machine learning or neural network algorithm comprising a regression block.