John Parker Burg
John Burg has made many fundamental contributions to exploration geophysics in the field of signal processing. His earliest were the development of multichannel Wiener filtering, the applications of it to deghosting, pie-slice filtering, and the analysis of signals from mantle P-waves while working at Geophysical Services Inc. and TI in the early 1960s. In 1963, when applying prediction error filters (PEF) to seismic recordings and studying Toeplitz matrix equations, Burg realized the filters were minimum phase and he was awarded the first patent on deconvolution. His work on the properties of PEFs led to maximum entropy spectral analysis (MEM) in the mid 1960s, and then to the Burg technique to produce better autocorrelations to which MEM could apply. In the 1980s he extended MEM to multiple dimensions. MEM has widespread use to this day, having found application not only in exploration geophysics but also in such diverse areas as speech processing, image processing, and astronomy.
Biography Citation for the SEG 2018 Reginald Fessenden Award
by M. Lee Bell
John Burg received bachelor's degrees in physics and mathematics from the University of Texas in Austin in 1953. He then earned a master's degree in physics at the Massachusetts Institute of Technology in 1960. He began his career in signal processing at Geophysical Services Inc. (GSI) in Dallas in 1961. He became a principle investigator on a government contract to design and build a seismic array observatory to detect underground nuclear explosions and derived an algorithm for rejecting surface waves and enhancing P-waves using Wiener multichannel filters. In 1963, while applying prediction error filters (PEFs) to seismic recordings and studying Toeplitz matrix equations, John realized the filters were minimum phase, which the exploration world was seeking. This led to the application for the first deconvolution patent, which was awarded in 1967. He also worked on deghosting (with Schneider, 1964) and pie-slice filtering (with Embree, 1963).
In the mid 1960s, his work on the properties of PEFs led to maximum entropy spectral analysis (MEM), an improved method of estimating the frequency spectrum of time series from autocorrelation measurements. However, the usual lag product of calculating an autocorrelation had shortcomings, especially in small windows, so John came up with the Burg technique of analyzing time series so the result was an autocorrelation and the MEM could then produce reliable spectra. John used these methods on several projects for GSI involving arrays of seismometers.
Leaving GSI to pursue his doctorate at Stanford, John developed MEM more fully in his thesis. MEM began to have widespread use in a number of fields besides exploration, such as in speech and image processing and astronomy. In later years, he extended MEM to multidimensional data (1982–1988) and improved the autocorrelation estimation by the introduction of the structured covariance method, where the entire autocorrelation is analyzed as a single entity. Some of this work was done after John founded his first company, Time and Spacing Processing, but most was done after starting his second company, Entropic Geophysical Inc.
As mentioned earlier, John's influence on the global signal processing community far transcends exploration geophysics. There are very few geophysicists whose work remains as fresh and influential today as it was almost half a century ago, and John is one of them.