Lower Limb Analysis Based on Surface Electromyography (sEMG) Using Different Time-frequency Representation Techniques

Using time-frequency representation techniques, projecting 1D sEMG signals onto a 2D image space can help diagnose several muscle activities. The acquired signal provide valuable representative information about the activity firing rates during contraction. Different phases of be discernible via by extracting discriminating features. behavior was in measurements five muscles, i.e., RF, BF, VM, ST, and FX. Previous attempts to visualize lower limb analysis extract features adopted One-dimensional (1D) segments. This work proposes comparative experiment between three techniques. scalogram, spectrogram, persistence spectrum, were used map muscles' straightening knee. two-dimensional (2D) projected images are then fed into convolutional neural network (CNN) model for detecting knee abnormality. experiments performed 10-fold cross-validation. number kernels is incremented along with layers. fully connected layers adjusted according loss value. Besides, tuning hyper-parameters dropout parameters ReLU activation function verify optimal performance. research shows that scalogram gives significantly better performance than spectrogram spectrum recognizing In addition, this study may guiding diagnosis human activities signal. A more diverse muscles further investigated useful future enhance accuracy.