EPDF: A Dual-Channel Individual Classification and Authentication Method Based on Feature Fusion of ECG and PPG Signals in Wearable Devices

With the advancement of wearable devices and biosensor technology, ECG and PPG are becoming prominent for individual classification and authentication due to their unique and hard-to-forge features. However, multimodal ECG and PPG authentication in wearables face issues like signal interference and power consumption. We propose “A Dual-Channel Individual Classification and Authentication Method Based on Feature Fusion of ECG and PPG Signals in Wearable Devices” (EPDF). EPDF uses a dual-channel approach with MultiRes blocks and Spatial Pyramid Pooling (SPP) for improved feature extraction and incorporates a self-attention mechanism and multi-channel cascading to reduce information loss. Experimental results show EPDF significantly improves FAR and EER compared to traditional methods, making it a suitable choice for wearable device classification.