EDA Feature Optimization for Tiny Wearable Stress Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing wearable sensor-based stress detection systems face challenges in achieving a balance between classification accuracy and computational resource efficiency, particularly on battery-powered devices, with manual design processes being time-consuming and inefficient.

Innovation Solution

Employing a Neural Architecture Search (NAS) technique to automate the generation of lightweight classification models for affective parameter determination from electrodermal activity, optimizing feature selection and model generation for resource-constrained devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual design process is used for ML/DL models, then classification accuracy can be optimized, but computational resource efficiency deteriorates and design time increases

Engineering Contradiction:
Improveclassification accuracyVSAvoidcomputational resource efficiency
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system automatically adjusts model parameters including architecture configuration, hyperparameters, and feature selection to find the optimal balance between accuracy and resource efficiency. The automated design process explores parameter spaces to identify configurations that minimize computational overhead while maintaining classification performance.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system performs self-optimization by automatically generating and evaluating multiple model configurations without manual intervention. The automated design process includes self-assessment of resource requirements and self-adjustment of model complexity to achieve the desired accuracy-resource efficiency tradeoff.

Inventive Principle:
Principle #25Self-service

2Measurement precision

If complex ML/DL models are deployed on wearable devices, then classification accuracy improves, but device complexity and resource consumption increase

Engineering Contradiction:
Improveclassification accuracyVSAvoidmodel complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the model design process into distinct components: feature extraction, architecture selection, hyperparameter optimization, and resource constraint evaluation. This segmentation allows for systematic optimization of each component to reduce overall model complexity while preserving accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically adjusts model complexity based on device-specific resource constraints. The automated design process adapts model architecture and parameters to match the computational capabilities of the target wearable device, creating dynamically optimized models rather than static complex architectures.

Inventive Principle:
Principle #15Dynamics

3Productivity

If automated model generation is implemented, then design time is reduced, but computational overhead during model generation increases

Engineering Contradiction:
Improvemodel generation speedVSAvoidcomputational overhead
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-defining search spaces, constraint boundaries, and evaluation metrics before the automated design process begins. This preliminary setup reduces the computational overhead during actual model generation by avoiding exploration of invalid or suboptimal configurations.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4614516A1Feature optimization and automated generation of tiny models for affective parameter determination from electrodermal activity
Publication Date: 2025.09.10 TATA CONSULTANCY SERVICES LTD
  • EP4614516A1 patent drawingFigure 1
  • EP4614516A1 patent drawingFigure 2
  • EP4614516A1 patent drawingFigure 3

AI summary

Wearable sensor-based stress detection is a well explored area of research in affective computing domain and are performed with non-invasive sensing modalities like Electrodermal Activity (EDA). In recent years, with the increased availability of such wearable devices to end-users, these applications have become more pervasive and thus require a greater level of optimization for continuous usage on resource constrained and battery-powered devices. While several research works have focused on designing Machine Learning (ML) and Deep Learning (DL) models for these tasks, very few focus on resource footprint. The balance between classification accuracy and computational resources is difficult to achieve with manual design process and is time consuming. In the present disclosure, systems and methods apply Neural Network Search Space (NASS) technique on features for feature optimization, and generation of tiny models based on optimized features set suitable for round-the-clock inference, with minimal resource requirements, and low latency.