System and Method for Detecting Person Activity in Video

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

This invention describes a system and method for automatically detecting and recognizing human activities in video footage. It utilizes machine learning to identify people in each video frame, extract person-specific features, analyze the surrounding scene, and determine the activities of each detected person. The system models both individual actions and interactions between people, enabling more nuanced activity recognition without requiring densely annotated training data.

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

• Surveillance systems for detecting suspicious or criminal activities in real time
• Behavior analysis in public areas (e.g., airports, malls) for safety and crowd management
• Automated video content tagging, indexing, and summarization
• Sports analytics to track individual and team actions during games
• Human-computer interaction systems that respond to user activity
• Elderly care monitoring to detect falls or unusual inactivity
• Retail analytics for understanding customer behavior
• Entertainment and video editing tools for automatic scene and action recognition

BenefitsContent extracted from patent full text and abstract with AI.

• Reduces the need for fully-annotated training datasets, lowering implementation costs
• Recognizes both individual and interactive group activities, increasing accuracy and context awareness
• Enables real-time analysis and detection of multiple simultaneous actions within video frames
• Adaptive to scenes with multiple people, dynamic interactions, and various activity types
• Can be integrated into existing video processing and surveillance infrastructure
• Scalable for various domains: security, healthcare, retail, entertainment, and more
• Provides rich metadata for improved search, retrieval, and analytics of video content

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

A system and a method for detecting person activity in a video are disclosed. The method comprises detecting one or more persons in a frame of the video; generating a first set of feature vectors based respectively on a set of frames of the video; generating a second set of person-specific feature vectors based on the set of frames; determining a temporal scene context vector for the frame, based on the first set of feature vectors and the second set of person-specific feature vectors; and for a detected person of the one or more detected persons, determining a hidden state vector for the frame, based on the temporal scene context vector and a subset of the second set of person-specific feature vectors corresponding to the detected person; and detecting one or more activities performed by the detected person in the frame based on the determined hidden state vector.