Stage 1 (2017–2019)
Classic2D Representations & Shallow CNNs
Time–Frequency/Wavelet/Heartbeat Reconstructi Classic CNN
Xia Y (2017) — SWT-CNN (2Conv + 2Pool + FC)
Jun TJ (2018) — VGG-variant (4ConvBlocks + 3Pool + 2FC)
Wu et al. (2019) — Custom 2D-CNN (4ConvBlocks + 2Pool + FC)
Tadesse GA (2019) — GoogLeNet Transfer (Inception + FC)
Early Residual Structure Attempts
Kachuee M (2018) — Transferable ResCNN (5Residual + GAP + FC)
Brisk et al. (2019) — ResNet Transfer (Conv + Pool + 2FC)
Stage 2 (2020–2022) DeepResidual Networks &Attention Enhancement
Classical Backbones + Input Diversification
Ullah et al. (2020) — 2D-CNN (4Conv + 4Pool + FC)
Mathunjwa et al. (2021) — AlexNet / VGG16 / VGG19 (RP images)
Jeong & Lim (2021) — STFT + CNN (4ConvModules + BN + FC)
Degirmenci et al. (2022) — Lightweight CNN (3Conv + 3Pool + FC)
Residual + Attention
Li et al. (2022) — SE-ResNet + S-shape (17Conv + 8Residual + SE + GAP)
Mathunjwa et al. (2022) — Multi-stage ResNet (ResNet-18 → deeper ResNet)
Multi-Lead / Multi-Branch Fusion
Kwon et al. (2020) — Residual CNN + MLP (multi-lead spatial–temporal fusion)
Stage 3 (2023–2025)Lightweight, Multi-Branch, Cross-Modal
Lightweight Multi-Branch CNN
Zhang et al. (2023) — ParNet-adv (triple-branch lightweight CNN)
Narotamo et al. (2024) — Multi-backbone comparison (AlexNet / VGG / ResNet / MobileNetV2 / AlexNetAtt)
Elyamani et al. (2024) — Separable Conv + Residual Blocks
Nonlinear Input Mapping
Elmir et al. (2023) — GAF-based images + CNN / VGG / ResNet / EfficientNet
Cross-Modal & Temporal Fusion
Kolhar & Al Rajeh (2024) — CNN + LSTM Hybrid (feature + temporal modeling)
Next-Gen CNN Backbones
Yoon & Joo (2025) — ConvNeXt (cross-scale convolutional backbone)
U668519106
U554883986
