Quantum enhanced EEG classifier towards brain-controlled wheelchair navigation

Brain-computer interfaces (BCIs) provide a pathway to assistive technologies such as brain-controlled wheelchairs, yet accurate motor imagery (MI) classification from electroencephalography (EEG) remains challenging due to noise and subject variability. In this work, we propose a hybrid Quantum Enhanced CNN-LSTM model EEG Classifier (HQeCL), incorporating a simulated quantum pooling layer for richer feature abstraction. The framework integrates power spectral density (PSD) from the frequency domain, common spatial patterns (CSP) from the spatial domain, and quantum entropy from the non-linear domain to capture complementary EEG characteristics. The model was evaluated using leave-one-subject-out (LOSO) cross-validation on the 8-channel motor imagery dataset, achieving 92.1%±5.9 accuracy, 93.1%±6.2 precision, 91.9%±1.3 recall, 92.5%±1.3 F1-score, and Cohen's κ=0.89±0.02. Compared to existing methods, HQeCL outperformed CSP-LDA (74.5%±1.4), ShallowConvNet (83.3%±1.6), and CNN-LSTM (88.8%±1.2), while remaining competitive with QuEEGNet (91.4%±1.3). Ablation analysis confirmed the contribution of quantum pooling, which provided a +0.7% gain over average pooling, and UMAP, which improved performance by +14.8% over PCA and +29.7% over t-SNE. Complexity analysis further demonstrated the efficiency of HQeCL with only 0.12M parameters, 270.2M FLOPs, and an inference latency of 77.6ms. While these results demonstrate near real-time feasibility in simulation, translation to hardware remains a challenge, positioning HQeCL as a quantum-inspired, Pareto-efficient EEG classifier advancing motor imagery decoding for brain-controlled wheelchair navigation.