Least-Squares System Design

The least-squares principle is widely applicable to the design of digital signal processing (DSP) systems. This chapter aims to extend the least-squares concept to the design of other kinds of signal processing systems. It shows how several kinds of DSP systems may be realized using a common linear least-squares approach. The chapter describes briefly a variety of tasks involving prediction, modeling, equalization, and interference canceling, where least-squares design is useful. It suggets that all these tasks lead to the same least-squares design problem, which is to match a particular signal to a desired signal so that the difference between the two signals is minimal in the least-squares sense. The chapter also describes system configurations where the least-squares design is applicable. The linear predictor is useful in waveform encoding and data compression, spectral estimation, spectral line enhancement, event detection, and other areas.