This study investigates demographic bias in ECG-based biometric recognition systems and proposes a data augmentation approach to mitigate performance disparities across gender, age, and smoking-status subgroups. Our analysis reveals significant accuracy gaps, with male subjects outperforming females by 18% and middle-aged adults (35–44 years) achieving 67% accuracy compared to just 14% for older adults (45–50 years). To address these disparities, we introduce a demographic-aware augmentation strategy that balances representation by oversampling underrepresented groups while preserving biometric feature integrity. The augmented model demonstrates a 28% improvement in overall accuracy (from 38% to 49%) while still reducing inter-group performance disparities. Notably, the originally worst-performing age group (45–50 years) became the highest-performing (66% accuracy) post-augmentation. However, persistent gaps in female recognition (40% vs. male 56%) and low F1 scores (0.26–0.39) suggest that some biases require additional mitigation strategies beyond data augmentation. These findings highlight the dual benefits of demographic balancing—enhancing both fairness and overall system performance—while underscoring the need for hybrid approaches combining augmentation with adversarial learning or attention mechanisms. This study provides actionable insights for developing more equitable biometric systems and establishes a framework for evaluating demographic bias in ECG-based recognition.