Wi-Fi CSI Viewer Presence Detection with Self-Training Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audience measurement methods for TV viewership rely solely on user input from panelists, which is limited and requires extrapolation to estimate viewership for larger populations, and do not efficiently utilize Wi-Fi Channel State Information (CSI) for accurate viewer presence detection.
Innovation Solution
A system and method for training Wi-Fi CSI machine-learned models using data from streaming devices and user inputs to predict viewer presence and number, incorporating continuous training and updating with anonymized and aggregated data from panelist devices, and utilizing machine-learned models like neural networks to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual user input from panelists is used for audience measurement, then data collection is simple to implement, but measurement precision and reliability are insufficient due to limited panelist participation and extrapolation requirements
Solution Approach 1:
The patent replaces manual user input (mechanical data collection method) with automated Wi-Fi CSI-based detection. The system uses wireless channel state information from Wi-Fi signals to automatically detect viewer presence and count viewers, eliminating the need for manual panelist reporting and significantly improving measurement precision while maintaining manageable system complexity through automated processing.
Solution Approach 2:
The patent introduces Wi-Fi CSI data as an intermediary medium to infer viewer presence. Instead of directly asking users to report their presence, the system uses Wi-Fi channel state information as an indirect indicator that correlates with viewer presence, enabling automated and accurate detection without requiring direct user input.
2Measurement precision
If Wi-Fi CSI data is collected and processed through machine-learned models, then measurement precision improves, but device complexity and data processing requirements increase
Solution Approach 1:
The patent performs preliminary actions by collecting and storing Wi-Fi CSI data over time before actual viewer detection is needed. The system continuously gathers channel state information and uses it to train machine-learned models in advance, so that when viewer presence detection is required, the models are already prepared and can provide accurate results without requiring complex real-time processing.
Solution Approach 2:
The patent implements self-service through automated model training and updating. The system automatically uses collected Wi-Fi CSI data to train and refine its own machine-learned models without requiring external intervention. This self-training mechanism improves measurement precision while managing complexity by automating the model development process.
3Reliability
If machine-learned models are continuously updated with training data, then reliability and accuracy improve over time, but loss of time and computational resources increase
Solution Approach 1:
The patent implements periodic action by updating machine-learned models at scheduled intervals rather than continuously. The system collects Wi-Fi CSI data over time and periodically retrains models with accumulated data, balancing reliability improvement with time efficiency. This approach allows the system to maintain accurate models while avoiding excessive time consumption associated with continuous retraining.
Data Source
AI summary
Example embodiments of the present disclosure provide for an example method for machine learned models for detecting viewer presence using Wi-Fi Channel State Information (CSI). For instance, the method can include obtaining data comprising Wi-Fi CSI from a receiver associated with a device. The method can include determining a predicted number of viewers in front of the device using a Wi-Fi CSI machine-learned model. The method can include displaying a prompt for a user to respond to via a second user computing device responsive to determining the predicted number of viewers. The method can include generating a training dataset based on the obtained user input, obtained Wi-Fi CSI, and the predicted number of viewers. Updating the Wi-Fi CSI machine-learned model based on the training dataset and using the updated Wi-Fi CSI machine-learned model to predict a number of viewers based on second Wi-Fi CSI data.