Title :
Autoregressive equivalents of rank order processors
Author_Institution :
Dept. of Electr. Eng. & Comput. Sci., Northwestern Univ., Evanston, IL, USA
fDate :
6/1/1990 12:00:00 AM
Abstract :
A theory is presented which shows that any rank order processor (ROP) specified as a positive regular set over the binary alphabet, and any ROP implementable via a stack encoding as input to identical finite-set sequential machines, is also implementable in finite autoregressive form with conventional input encoding, provided an internal state vector is properly defined. Any such ROP algorithm can be factored into computations of the median of three values. The theory also shows how ROPs may be derived from arbitrary partially ordered finite sets
Keywords :
filtering and prediction theory; finite automata; set theory; signal processing; time series; binary alphabet; finite autoregressive form; finite-set sequential machines; internal state vector; partially ordered finite sets; positive regular set; rank order processors; signal processor; stack encoding; Acoustics; Automata; Encoding; Filters; Lattices; Mathematical model; Polynomials; Signal processing; Signal processing algorithms; Speech;
Journal_Title :
Acoustics, Speech and Signal Processing, IEEE Transactions on
