Wireless Sensing User Identification Model Generation for Local Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless sensing-based user identification technologies face challenges in creating a generalized model due to varying signal patterns in different environments, making supervised learning impractical for commercial products as it requires user participation for data collection and labeling.
Innovation Solution
A post-learning method is proposed that collects personal identification information to tailor the learning process, allowing for unsupervised, supervised, or semi-supervised learning to create a user identification model suitable for each environment, enabling AI devices to identify individuals effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If supervised learning is used to create a generalized user identification model, then the model can be applied across different environments, but it requires user participation for data collection and labeling which reduces ease of operation
Solution Approach 1:
The system performs automatic device discovery and collects wireless sensing data without requiring user participation. The user identification device autonomously discovers PEDs, obtains their identification information, and collects labeled data through normal device operations, enabling the model training process to serve itself without external intervention.
Solution Approach 2:
The system collects and stores wireless sensing data and PED identification information in advance during normal device operations. This preliminary data collection occurs before model training is needed, so when training is required, the labeled data is already available, eliminating the need for user participation at the time of model creation.
2Ease of operation
If wireless sensing data is collected without PED identification information, then data collection is simpler, but the accuracy of user identification decreases due to lack of labeled data
Solution Approach 1:
The system merges the collection of wireless sensing data with the acquisition of PED identification information by utilizing existing discovery mechanisms. When the user identification device discovers a PED, it simultaneously obtains the PED's identification information and collects the corresponding wireless sensing data, combining two functions into one process without increasing operational complexity.
Solution Approach 2:
The PED identification information acts as an intermediary that links wireless sensing data to specific users. By using the PED ID as a mediator, the system can automatically label collected data without direct user involvement, maintaining data collection simplicity while enabling accurate user identification through supervised learning.
3Measurement precision
If a user identification model is trained for each specific environment, then identification accuracy improves, but the device complexity increases due to multiple environment-specific models
Solution Approach 1:
The system adapts to different environments by changing the training parameters and data characteristics rather than creating separate model structures. By collecting environment-specific labeled data and retraining the same model architecture with environment-appropriate parameters, the system achieves high accuracy in each environment without increasing structural complexity.
Solution Approach 2:
The system dynamically adapts to different environments through post-learning processes. When deployed in a new environment, the device automatically collects local data, obtains PED identification information, and retrains the model to adapt to the specific environment's characteristics, making the system flexible rather than requiring static pre-deployment configuration.
Data Source
AI summary
Proposed is a method and device for generating a user identification model on the basis of wireless sensing in a wireless LAN system. Specifically, a user identification apparatus discovers a PED and acquires identification information about the PED. The user identification apparatus collects data and pre-processes the collected data on the basis of the identification information about the PED. The user identification apparatus generates a user identification model by learning the pre-processed data.