An analytical approach to signal reconstruction using Gaussian approximations applied to randomly generated and flow cytometric data

Abstract

This study introduces an analytical approach to signal reconstruction using Gaussian distributions. A major problem encountered in real-world data distributions is in the ability to accurately separate those data distributions that experience overlap. A first objective then is to develop a method of determining accurately the characteristics of a given distribution even when it has been affected by another distribution that lies close to it. In addition, normally, two-dimensional (2-D) Gaussian distributions are described by means of a correlation coefficient, but in this case, a normal 2-D distribution will be assumed in a direction parallel to a reference axis and then rotated by some angle /spl theta/. This outcome will not affect the results in terms of the standard use of the correlation coefficient. In this study, an attempt is made to provide a highly accurate yet computationally inexpensive approach of resolving the problem of overlap as we seek the reconstruction of signals through Gaussian curve fitting. Implementation results are shown in support of this assertion.