User Pose Matching for Accurate Identification in Crowds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user identification technologies face challenges in accurately identifying a specific user in crowded environments without requiring pre-established characteristic information and high-resolution cameras, especially in scenarios like calling a taxi during peak hours.
Innovation Solution
A user identification device and method that utilizes a transceiver interface and processor to receive device pose information from a target device, analyze real-time images to calculate human body pose information, and compare it with device pose information to identify the target user, using inertial sensors and image capturing devices without the need for high-resolution cameras.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image recognition technology with pre-established characteristic information and high-resolution camera is used, then user identification accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The patent extracts only the necessary pose information (position and orientation) from the full image data, rather than performing complete image recognition. The image capturing device captures images, but only pose parameters are extracted and transmitted to the server, reducing the complexity of on-device processing while maintaining identification accuracy.
Solution Approach 2:
Instead of having the user's device perform complex image recognition and feature extraction, the patent inverts the approach by having the server perform the analysis. The user's device only needs to capture images and transmit pose information, while the heavy computational task of identification is moved to the server side.
2Device complexity
If GPS location is used for user identification, then device complexity is reduced, but identification accuracy deteriorates due to several meters error
Solution Approach 1:
The patent merges GPS location data with pose information (position and orientation from image analysis) to create a more precise identification system. By combining these two types of data, the system achieves better accuracy than GPS alone while keeping device complexity manageable.
Solution Approach 2:
The patent adds the dimension of pose information (orientation and position relative to the image) to the traditional GPS location data. This transforms the identification from purely location-based to a multi-dimensional approach that includes spatial orientation, thereby improving precision without significantly increasing device complexity.
3Measurement precision
If feature comparison algorithms are executed, then user identification accuracy is improved, but computational cost increases
Solution Approach 1:
The patent extracts only pose parameters (position and orientation) from the captured images, rather than performing full feature extraction and comparison. This extraction approach significantly reduces computational cost while maintaining sufficient accuracy for identification purposes.
Solution Approach 2:
The patent inverts the computational burden by moving the feature comparison and identification algorithms from the user's device to the server. The device only performs lightweight image capture and pose parameter extraction, while the energy-intensive computational tasks are executed remotely.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A user identification device, method, and non-transitory computer readable storage medium thereof are provided. The device receives a device pose information from a target device, the target device is used by a target user, and the device pose information includes a three-degree-of-freedom pose information of the target device. The device analyzes a real-time image including a plurality of users to be identified to calculate a human body pose information corresponding to each of the plurality of users to be identified. The device compares the plurality of human body pose information and the device pose information to identify the target user from the plurality of users to be identified.