Privacy-Preserving Fall Detection Using Local Vision and Cloud Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional video surveillance technologies fail to protect user privacy and are resource-intensive, limiting their use in private settings and increasing storage and bandwidth costs.
Innovation Solution
A system that uses a local vision sensor to extract privacy-preserving human representations and background images, transmitting only skeleton sequences to a cloud server for complex action recognition, reducing data size and preserving privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional video surveillance technologies are used to monitor private areas, then fall detection capability is improved, but user privacy is compromised
Solution Approach 1:
The patent extracts only the essential motion information (skeleton sequences) from the video data while removing all personally identifiable visual information. The local device processes video frames to generate skeleton sequences representing human pose and motion, which are then transmitted to the cloud server for fall detection analysis, leaving the original video data stored only locally.
Solution Approach 2:
The system segments the video processing workflow into two distinct parts: local processing that extracts skeleton sequences from video frames, and cloud-based processing that analyzes these sequences for fall detection. This segmentation allows privacy-sensitive video data to remain local while enabling remote analytical capabilities.
2Adaptability or versatility
If recorded videos are transmitted and stored in the cloud for future analysis, then video content retrieval and analysis capability is improved, but network bandwidth and storage space requirements increase significantly
Solution Approach 1:
The patent extracts only the essential motion information (skeleton sequences) from the video data while removing all personally identifiable visual information. The local device processes video frames to generate skeleton sequences representing human pose and motion, which are then transmitted to the cloud server for fall detection analysis, leaving the original video data stored only locally.
Solution Approach 2:
The system transforms video data from its original high-dimensional pixel format into a compressed skeleton sequence representation with significantly fewer parameters. Each skeleton sequence consists of a small number of keypoint coordinates and motion vectors, reducing data volume by orders of magnitude while preserving the essential motion information needed for fall detection.
3Quantity of substance
If skeleton sequences are transmitted to cloud server instead of full video, then network bandwidth and storage requirements are reduced, but processing complexity at local device increases
Solution Approach 1:
The local device performs preliminary processing by extracting skeleton sequences from video frames before transmission. This pre-processing step converts raw video data into a compressed representation that is easier to transmit and analyze, shifting the computational burden from the cloud server to the local device.
Solution Approach 2:
The patent replaces the mechanical transmission and storage of large-volume video data with a computational approach that generates compact skeleton sequence representations. This substitution uses algorithmic processing to transform visual data into a more efficient format for transmission and cloud-based analysis.
Data Source
AI summary
In one aspect, a fall detection system first receives, at a local vision sensor, a video image including at least a person. Further at the local sensor, the system detects the person in the video image, extracts a privacy-preserving human representation of the detected person and a background image from the video image, and classifies the extracted human representation of the detected person into an action among a set of predetermined actions which includes a fall alert. Next, at a cloud server, the system receives the human representation, the classified action, and the background image from the local vision sensor. Further at the server, the system processes the human representation and the background image to determine if a fall has occurred, and if not, further determine if the classified action is a fall alert. If so, the fall alert is ignored, thereby reducing a false alarm rate of fall detection.