Special-Purpose Orthonormal Basis Functions - Application to Motor Unit Action Potentials

Signal representation is a useful technique to reduce the dimensionality of biological potentials. Usually the basis components are well-defined mathematical functions. We present a new approach using a basis composed of a set of experimentally measured examples of the potential which is to be represented. The criteria used in selecting these basis potentials are described. Using this technique, motor unit potentials measured by a concentric electromyography needle are represented by a basis having 18 components. This technique was found to require roughly one-third fewer components than Fourier analysis to achieve the same quality of representation.