Server-Based User Data Annotation for Behavior Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The limited amount of experimental annotation data restricts the application of user behavior recognition, as it relies heavily on categorization of training samples, which is insufficient for accurate inference of user behaviors or intentions.
Innovation Solution
A method involving a server and terminal device that collects and annotates location and behavior information through a game process, using sensor data to generate richer annotation data by predicting candidate user location and behavior information and allowing user selection, thereby increasing the dataset for user behavior recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If experimental annotation data is used for user behavior recognition, then the categorization process can be performed, but the limited amount of data restricts the application and accuracy of user behavior recognition
Solution Approach 1:
The patent applies preliminary action by collecting and annotating user data in advance through game processes before actual user behavior recognition tasks. The system pre-collects location information, behavior information, and sensor data during game playing, and pre-annotates this data with user behavior labels, creating a ready-to-use training dataset that improves both data quantity and recognition accuracy without limiting application scope
Solution Approach 2:
The system implements self-service by automatically collecting sensor data from terminal devices and automatically annotating user behavior information without requiring extensive manual annotation efforts. The server autonomously processes the data collection, annotation, and storage processes, enabling the system to generate its own training data efficiently and scale the dataset without proportional increases in manual work
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of the present invention provide a user data annotation method, a terminal device, and a server. The method includes: receiving, by a server, location information of a first user corresponding to a first terminal device, behavior information of the first user, and sensor data of each sensor on the first terminal device, which are sent by the first terminal device in a process of a game; and annotating, by the server, the location information of the first user, the behavior information of the first user, and the sensor data to obtain annotation data of the first user. The technical solution of the present invention can increase the amount of annotation data and overcome the limitation on the application of user behavior recognition.