Tiny DQN Model Generation for Wearable Pain Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wearable devices lack an efficient and scalable method for real-time pain detection using multi-sensor fusion, requiring manual model design and lacking flexibility to adapt to different devices and use-case settings, while also failing to address the variability among individuals and the interconnectedness of pain and stress.
Innovation Solution
A system and method for generating tiny Deep Q-Network (DQN) models with an optimal set of sensors for wearable devices, utilizing an automation framework that intelligently selects and fuses sensor data based on domain knowledge and device constraints, enabling real-time pain detection and quantification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual model design and multi-sensor fusion are used for pain detection, then detection accuracy can be improved, but device complexity and computational requirements increase
Solution Approach 1:
The system employs automated framework that self-configures sensor selection and model architecture based on device constraints and pain type, eliminating the need for manual model design. The framework automatically generates optimized DQN models tailored to specific wearable devices and use-case settings, enabling the system to serve itself without expert intervention.
Solution Approach 2:
The system dynamically adjusts model parameters and sensor configurations based on device constraints and operational requirements. By changing parameters such as model size, sensor subset, and architecture configuration, the system achieves accurate pain detection while adapting to different computational resources and device capabilities.
2Measurement precision
If multiple sensors are used for comprehensive pain assessment, then measurement precision improves, but power consumption and computational load increase
Solution Approach 1:
The automated framework extracts and selects only the essential sensors and features needed for accurate pain detection on edge devices. By taking out unnecessary sensors and model components, the system reduces power consumption and computational load while maintaining detection accuracy through optimized sensor subsets and streamlined model architectures.
3Adaptability or versatility
If customized models are created for different devices and use-cases, then adaptability improves, but manufacturing precision and model optimization difficulty increase
Solution Approach 1:
The system implements a universal automated framework that can generate customized models for any wearable device and use-case configuration. This multi-functional framework handles sensor selection, model architecture design, and optimization automatically, making the deployment process easy across diverse devices while maintaining high adaptability to specific requirements.
4Speed
If real-time pain detection is implemented on edge devices, then response speed improves, but computational constraints and model size requirements worsen
Solution Approach 1:
The system segments the pain detection task into modular components: sensor data acquisition, feature extraction, and DQN-based prediction. By dividing the computational workload and using lightweight DQN models with optimized architectures, the system achieves real-time detection on resource-constrained edge devices while managing computational complexity through structured model design.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
State of art techniques hardly address real-time detection of pain with sensor fusion approaches. A method and system herein provides intelligent sensor selection and fusion mechanism to detect the presence of pain and quantify pain level on edge devices in an effective and efficient manner, guided by information on pain origin, pathway and type extracted from input signals, without the requirement of any manual intervention or interference. The system provides an automation framework to accelerate the deployment of customized tiny models for wearable and edge devices, based on sensor availability and device compute capacity to create suitable model for the target device and task. The system takes accuracy, model size and latency as objectives, and sensor availability and computation resource constraints as targets for the automation framework enabling rapid generation and deployment of models on multiple edge devices (wearable devices), independent of the availability of a specific sensor.