A neural network for noninvasive decomposition of surface EMG signals using Gaussian nodes

The decomposition of surface EMG (electromyograms) signals into their constituent single fiber action potentials (SFAPs) is addressed. Generally, this problem is analytically not tractable and is computationally too complex to be reliable. It is demonstrated that it can be resolved by a specially designed neural network called the neural network with Gaussian nodes. Together with a modified back-propagation algorithm, a method of choosing initial conditions is presented. The significance of such solutions is that they allow a physician or medical researcher to observe the time behavior of SFAPs in a noninvasive manner for diagnostic purposes or other medical applications.<
>