surface electromyographic data simulation
The model proposed by Farina et al was used to generate surface EMG signals [1]. In this model, the volume conductor was described as an anisotropic multilayered cylinder and the source was a spatio-temporal function describing the generation, propagation, and extinction of the intracellular action potential at the end-plate, along the fiber, and at the tendons, respectively. The Inter-Electrode-Distance (IED) was set to 5 mm as recommended in [2] to locate IZs. The remainder of the model parameters used in our study were in principle the same as those used by Mesin et al [3]. Finally, the number of active MUs in each 60-ms simulated signal interval was between 1 and 5. Signals were zero-phase digitally band-pass filtered [4] using an overall eighth-order Butterworth filter with cut-off frequencies 20 and 500 Hz.
For each MU number category (1 to 5), sEMG signals with SNR values of -5, 0, 5, 10 and 15 dB were simulated to include very low to moderate quality sEMG signals. Twenty Single-Differential (SD) channels were simulated along the muscle fiber direction and sampling frequency was 4096 Hz. Thirty frames (or images) with up to 5 IZs were simulated for each SNR value. The temporal location of the IZs was created randomly in each frame. The signal SNR for each simulated 60-ms epoch was defined as the RMS of the raw sEMG divided by the standard deviation of the added Gaussian noise, expressed in dB [5]. Thus, a total of 750 1-D linear array sEMG signals were simulated, considering five SNR values and maximum five MUs . We also provided the gold standard data for the IZ channels and CV values.