Abstract: The invention provides a method and apparatus for filtering a temporal signal. A target magnitude frequency response HT(f) is specified (101,201) of frequency f in terms of a column vector l of K weights lk where log HT(f)=lTW(f) and W(f) is a column vector of K magnitude basis functions Wk(f). A constrained frequency response Hc(f) is computed (102,214) defined by log Hc(f)=gTV(f) , where V(f) is a column vector of N constrained basis functions Vn(f) for which each exp gnVn(f) satisfies a constraint preserved by concatenation, and g is a column vector of N coefficients satisfying a matching criterion between lTW(f) and gTV(f). An input temporal signal is received (103,212) and filtered (104,210) with the constrained frequency response Hc(f) to form a filtered temporal signal; and the filtered temporal signal is output (105,211).

Abstract: The invention provides a method and apparatus for filtering a temporal signal. A target magnitude frequency response HT(f) is specified (101,201) of frequency f in terms of a column vector l of K weights lk where log HT(f)=lTW(f) and W(f) is a column vector of K magnitude basis functions Wk(f). A constrained frequency response Hc(f) is computed (102,214) defined by log Hc(f)=gTV(f) , where V(f) is a column vector of N constrained basis functions Vn(f) for which each exp gnVn(f) satisfies a constraint preserved by concatenation, and g is a column vector of N coefficients satisfying a matching criterion between lTW(f) and gTV(f). An input temporal signal is received (103,212) and filtered (104,210) with the constrained frequency response Hc(f) to form a filtered temporal signal; and the filtered temporal signal is output (105,211).