Algorithm Kings: The Birth of Digital Signal Processing

摘要

There will always be signals; they will always need to be processed; and there will always be new technologies and algorithms for implementing the processing. That conviction has given me a feeling of job security ever since I first fell in love with signal processing as an undergraduate student at MIT in the mid-1950s. Added to that is the fact that new mathematical insights and formalisms inevitably change the structure of algorithms and the types of processors used to implement them. The technology and mathematics for signal processing continue to evolve, and clearly the signal processing algorithms and platforms of today would have been viewed as witchcraft during my undergraduate years. Figure 1 is a photograph of a state-of-the-art spectrum analyzer used at Pratt and Whitney in the early 1960s for analyzing jet engine noise. The technology consisted of discrete components, including individually packaged transistors and probably vacuum tubes as well. In contrast, spectrum analyzers today are primarily software-based digital processors that take full advantage of the fast Fourier transform algorithm and other sophisticated signal processing algorithm techniques based on mathematical insights and using implementation technologies that were not available at the time the picture was taken.