DT4ECG: A Dual-Task Learning Framework for ECG-Based Human Identity Recognition and Human Activity Detection

TL;DR

DT4ECG tackles the problem of simultaneously identifying individuals and detecting activities from ECG signals. The authors propose a dual-task framework built on a 1D-CNN backbone with 1D residual modules, augmented by a novel Sequence Channel Attention (SCA) mechanism and GradNorm to balance multi-task gradients. Key contributions include the integration of SCA for joint channel and temporal feature refinement, the use of GradNorm to harmonize training across tasks, and validation on a custom multi-condition ECG dataset achieving 99.12% identity accuracy and 90.11% activity accuracy. The results suggest strong potential for ECG-based biometrics to enable secure, context-aware health and user experiences in wearables and smart devices.

Abstract

This article introduces DT4ECG, an innovative dual-task learning framework for Electrocardiogram (ECG)-based human identity recognition and activity detection. The framework employs a robust one-dimensional convolutional neural network (1D-CNN) backbone integrated with residual blocks to extract discriminative ECG features. To enhance feature representation, we propose a novel Sequence Channel Attention (SCA) mechanism, which combines channel-wise and sequential context attention to prioritize informative features across both temporal and channel dimensions. Furthermore, to address gradient imbalance in multi-task learning, we integrate GradNorm, a technique that dynamically adjusts loss weights based on gradient magnitudes, ensuring balanced training across tasks. Experimental results demonstrate the superior performance of our model, achieving accuracy rates of 99.12% in ID classification and 90.11% in activity classification. These findings underscore the potential of the DT4ECG framework in enhancing security and user experience across various applications such as fitness monitoring and personalized healthcare, thereby presenting a transformative approach to integrating ECG-based biometrics in everyday technologies.

Figures (7)

• Figure 1: The three-lead electrodes method to capture ECG signal
• Figure 2: The configurations of subject group
• Figure 3: The overview of 1D residual module
• Figure 4: The workflow of the SCA module. (a) Channel attention mechanism (b) Sequence attention mechanism
• Figure 5: The overview of the proposed DT4ECG framework