Confidence-driven Gradient Modulation for Multimodal Human Activity Recognition: A Dynamic Contrastive Dual-Path Learning Approach

TL;DR

This paper introduces a novel multimodal human activity recognition framework that dynamically aligns features across modalities using contrastive learning and confidence-driven gradient modulation, improving recognition accuracy and training stability.

Contribution

The proposed DCDP-HAR framework innovatively combines dual-path feature extraction, multi-stage contrastive learning, and dynamic gradient modulation to address cross-modal alignment and modality imbalance issues.

Findings

Effective cross-modal feature alignment demonstrated

Improved recognition accuracy on benchmark datasets

Enhanced training stability through gradient modulation

Abstract

Sensor-based Human Activity Recognition (HAR) is a core technology that enables intelligent systems to perceive and interact with their environment. However, multimodal HAR systems still encounter key challenges, such as difficulties in cross-modal feature alignment and imbalanced modality contributions. To address these issues, we propose a novel framework called the Dynamic Contrastive Dual-Path Network (DCDP-HAR). The framework comprises three key components. First, a dual-path feature extraction architecture is employed, where ResNet and DenseNet branches collaboratively process multimodal sensor data. Second, a multi-stage contrastive learning mechanism is introduced to achieve progressive alignment from local perception to semantic abstraction. Third, we present a confidence-driven gradient modulation strategy that dynamically monitors and adjusts the learning intensity of each modality branch during backpropagation, effectively alleviating modality competition. In addition, a momentum-based gradient accumulation strategy is adopted to enhance training stability. We conduct ablation studies to validate the effectiveness of each component and perform extensive comparative experiments on four public benchmark datasets.