Idea
Real-time cognitive load assessment platform from eye-gaze data for adaptive safety-critical human-centered AI applications.
Research Paper
Core Innovation
This paper introduces MambaGaze, which explicitly models missing eye-tracking data uncertainty via XMD encoding and captures long-range temporal dependencies with a bidirectional Mamba-2 model of linear complexity. This approach outperforms existing CNN, Transformer, ResNet, and VGG baselines in accuracy and supports efficient real-time edge deployment.
Why It Matters
Accurate real-time cognitive load monitoring enables adaptive systems in safety-critical domains like driver vigilance and flight assistance, improving safety and performance. Handling missing eye-tracking data and efficient temporal modeling enhances reliability and scalability for wearable and edge devices.
Market Size (TAM)
$2–10B TAM for cognitive load and human state monitoring platforms; $500M–$1B SAM from automotive, aviation, and wearable device sectors. Driven by increasing demand for safety-critical adaptive AI and wearable health monitoring.
Potential Customers & Pain Points
- Automotive manufacturers – Need reliable driver vigilance monitoring
- Aviation companies – Require adaptive flight deck assistance
- Wearable device makers – Demand low-power real-time cognitive load sensing
- AI safety system developers – Need robust human state assessment under data loss.
Business Model
Licensing the MambaGaze cognitive load assessment SDK to automotive, aviation, and wearable device manufacturers; offering edge-optimized inference software and consulting for integration.
Competitive Landscape
- Tobii
- Seeing Machines
- Smart Eye
- Affectiva
Implementation Challenges
- Integration with diverse eye-tracking hardware
- Regulatory approvals for safety-critical applications
- User privacy and data security concerns
Validation Strategy
- Conduct pilot deployments with automotive and aviation partners for driver and pilot monitoring
- Benchmark real-time performance and accuracy on diverse eye-tracking hardware
- Obtain regulatory feedback and certifications for safety-critical use cases
Research Paper Overview
MambaGaze: Bidirectional Mamba with Explicit Missing Data Modeling for Cognitive Load Assessment from Eye-Gaze Tracking Data
Summary
MambaGaze is a cognitive load assessment framework using eye-tracking data that handles missing data and models long-range temporal dependencies efficiently. It achieves higher accuracy than CNN, Transformer, ResNet, and VGG baselines on benchmark datasets and supports real-time edge deployment with low power consumption.