Maha Alsayyari

Computer Science, W_{j}=X[t_{j}:t_{j}+S-1],\quad t_{j+1}=t_{j}+S/2 ⊳\triangleright sliding windows 9: yj=arg⁡maxℓ∈ℒ⁡count⁡{yt=ℓ,t∈[tj,tj+S−1]}y_{j}=\arg\max_{\ell\in\mathcal{L}}\;\operatorname{count}\{y_{t}=\ell,\ t\in[t_{j},t_{j}+S-1]\} ⊳\triangleright majority label 10: 𝒟balancedtrain=OVRS⁡(𝒟train)\mathcal{D}^{\text{train}}_{\text{balanced}}=\operatorname{OVRS}(\mathcal{D}^{\text{train}}) ⊳\triangleright oversampling TABLE I: Deep learning models evaluated for EEG classification, grouped by architecture family. Family Model Purpose Focus Strengths CNN EEGNet [18] Compact EEG decoding Spatial-temporal features Robust to noise; lightweight DeepConvNet [33] Deep feature extraction Spatial-temporal features Strong representation learning ShallowConvNet [33] Short-window EEG decoding Temporal patterns Interpretable filters; effective for MI STNet [41] Multi-channel EEG modelling Spatial-temporal dependencies Captures electrode interactions TSCeption [8] Frequency-aware EEG decoding Temporal-spectral features Multi-scale kernels CCNN [37] Real-time EEG classification Low-latency processing Lightweight; portable BCI suitability CNN, WD Rover Pro (Rover Robotics), equipped with the following sensors mounted on a custom payload to enable reliable multimodal data acquisition in outdoor environments (see Figure 1): • Stereo Camera: StereoLabs ZED stereo vision camera providing real-time RGB and depth sensing, operating at 15 Hz, with a maximum field of view of 90° (H) × 60° (V) × 100° (D) and depth range of 0.5 m to 25 m. • GNSS: Yoctopuce Yocto-GPS-V2 module based on the u-blox NEO-M