Regression filters operate on the assumption that the slowly varying clutter component in the signal can be approximated by a polynomial [HdVD$^+$95]. The least-squares fit to the low-frequency clutter component in the echo signal is then subtracted from the signal. Regression filters therefore work on a different concept compared to FIR or IIR filters, which are based on theories that signals are the superposition of sinusiods. Thus, the regression filter design is not based upon commonly known impulse or frequency response concepts. In order to compare the regression filters to traditional filters (IIR, FIR), a frequency response can be calculated with Eq. 4.5. The polynomials can be chosen to form an orthonormal basis for a $K$-dimensional clutter subspace of the $N$-dimensional signal space. The least$
发表于 2010-2-13 01:36
提升卡
变色卡
显身卡
楼主