Continuous prediction of finger joint angles based on time series feature fusion CNN

Continuous Motion Intention Prediction Using sEMG for Upper-Limb Rehabilitation: A Systematic Review of Model-Based and Model-Free Approaches

“…The DL networks used in research can be classified into three categories: CNNs, RNNs, and CNN-RNN hybrids. The CNNs include one-dimensional CNN (1D-CNN) [124], 2D-CNN [125], 3D-CNN [126], [127], parallel CNN [128]- [130], AlexNet and ResNet [131], [132], as well as Temporal Convolutional Networks (TCN) [133], and Large-Scale TCN (LS-TCN) [134]. The RNNs encompass NARX [135], LSTM [13], [131], [132], [136]- [139], and GRU [131], [132].…”

“…CNNs: According to [124], 1D-CNN exhibited superior realtime prediction performance for finger force compared to 2D-CNN and LR due to its ability to learn deeper advanced features while reducing data dimensions and avoiding redundant spatial information. [125] suggested that using TD and FD feature images as inputs to 2D-CNN can further reduce noise and improve predictive accuracy compared to raw sEMG images. [126], [127] demonstrated that 3D-CNN can learn deeper muscle anatomy, MS, and motion velocity features from multiple electrode perspectives, enabling the prediction of untrained random new movements.…”