Behavior Event Data Transmission for Video Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing person detection systems require continuous monitoring of video data for safety assessment, leading to privacy concerns and high bandwidth usage, and struggle to efficiently detect and record behavior information from video content.
Innovation Solution
A recognition data transmission device that detects individuals and their behaviors from video, transmitting information in units of behavior events, utilizing image capturing, person detection, motion detection, behavior recognition, data conversion, and transmission control units to reduce data transmission and enhance convenience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If video data is transmitted continuously for person monitoring, then the watching side can continuously monitor person behavior, but the data transmission bandwidth becomes enormous and privacy is not protected
Solution Approach 1:
The patent extracts only the essential behavior information from complete video data. The behavior recognition unit identifies specific behaviors (e.g., walking, sitting, standing) and transmits only these extracted behavior events along with timestamps, rather than transmitting the entire video stream. This extraction principle reduces data volume while maintaining monitoring effectiveness.
Solution Approach 2:
The patent segments continuous video monitoring into discrete behavior events. The behavior recognition unit divides the continuous video stream into separate behavior instances, each with start time, end time, and behavior type. This segmentation allows transmission of individual behavior events rather than continuous video, reducing bandwidth requirements while preserving monitoring capability.
2Loss of information
If representative thumbnails are recorded at fixed intervals, then face detection and search are enabled, but detection information increases due to repeated recording of the same person
Solution Approach 1:
The patent uses feedback mechanisms to avoid redundant recording. The behavior recognition unit continuously monitors video input and compares detected persons with previously recorded behavior events. When the same person is detected again, the system checks whether a new behavior event has occurred. Only when behavior changes does the system record new data, using feedback from previous recordings to prevent duplication.
Solution Approach 2:
The patent transitions from static fixed-interval thumbnail recording to dynamic behavior-event-driven recording. Instead of recording at predetermined time intervals, the system dynamically adjusts recording based on detected behavior events. Recording occurs only when behavior changes are detected, making the recording frequency adaptive rather than fixed, thereby reducing redundant data storage.
3Loss of information
If multiple representative thumbnails are recorded within one behavior time, then comprehensive coverage is achieved, but the relation among detection information becomes unclear and behavior connection is difficult
Solution Approach 1:
The patent segments behavior monitoring into distinct behavior events with clear temporal boundaries. Each behavior event is assigned a unique identifier with start time and end time, creating discrete segments rather than overlapping thumbnails. This segmentation clearly delineates the temporal and logical relationships between different behaviors, making it easy to track and analyze behavior sequences without ambiguity.
Data Source
AI summary
A feature of a part or the entirety of a body of a person is detected from a captured video; the person in the video is specified from the detected feature; user information indicating physical features of the person is detected from the feature related to the specified person; motion information including a motion or a gesture with a body and hands of a user is detected from the user information and the captured video; a behavior including the motion of the person is recognized from the motion information and the user information; recognized behavior information is divided for each behavior of the person; the divided behavior information is generated as block data; and the generated block data is transmitted to an outside for each block.