Attention Estimation Using Trainer Feedback and Physical States
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for estimating user attention, such as those described in Patent Document 1, fail to accurately consider the intensity and spatial extent of attention due to limitations in accounting for user perception of changes in displayed information.
Innovation Solution
An information processing apparatus and method that utilizes a deep reinforcement learning model to estimate user attention based on physical states and experiences with changes in output information, determining the degree of change in output mode through feedback from a trainer, allowing for precise estimation of attention intensity and spatial extent.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If eye movement changes are used to estimate attention level, then attention estimation is achieved, but intensity and spatial extent of attention cannot be sufficiently considered
Solution Approach 1:
The patent segments the attention estimation into multiple dimensions: attention level (binary detection), attention intensity (degree of change in output mode), and spatial extent (location of attention). This segmentation allows each aspect to be measured and processed separately, preventing information loss about intensity and spatial characteristics while maintaining overall estimation accuracy.
Solution Approach 2:
The patent adds new dimensions to the attention estimation by incorporating trainer physical states (posture, gestures, facial expressions) and output mode changes as additional measurement axes. This transforms the estimation from a single-dimensional eye movement metric to a multi-dimensional assessment that captures intensity and spatial extent of attention.
2Measurement precision
If trainer physical state and experience are incorporated into the estimation model, then attention intensity and spatial extent are accurately captured, but system complexity increases
Solution Approach 1:
The patent implements feedback loops where trainer physical states and experience data are continuously fed into the estimation model, which then adjusts its predictions. This feedback mechanism allows the system to learn from trainer responses and improve accuracy over time without requiring complete redesign of the entire system architecture.
Solution Approach 2:
The estimation model performs self-adjustment by automatically incorporating trainer physical state data and experience information to refine its own predictions. The system serves itself by using its own output (attention estimates) to generate new input data (trainer responses), creating a self-improving cycle that reduces the need for external calibration.
3Measurement precision
If deep reinforcement learning is used to determine degree of change in output mode, then precise attention estimation is achieved, but computational requirements and processing time increase
Solution Approach 1:
The patent performs preliminary actions by pre-training the deep reinforcement learning model with extensive trainer data and experience information before actual attention estimation begins. This pre-processing creates a ready-to-use model that can make rapid predictions during actual use, transferring the computational burden from real-time operation to offline preparation.
Solution Approach 2:
The patent optimizes model parameters and thresholds based on pre-collected data, adjusting sensitivity and decision boundaries to achieve accurate estimation with minimal processing. By tuning parameters like attention thresholds and output mode change sensitivity during setup, the system reduces computational requirements during actual runtime while maintaining precision.
Data Source
AI summary
[Problem] To accurately estimate intensity of a user's attention and a spatial extent of the attention.[Solution] An information processing apparatus includes an estimation unit that estimates a level of attention of a user to output information of a certain type on a basis of a physical state of the user. The estimation unit estimates the level of attention using an estimation model that determines a degree of change in an output mode of the output information on a basis of a physical state of a trainer and an experience of whether or not the trainer has perceived a change in an output mode of training information of the same type as that of the output information in a case where the output mode has been changed.