Generalized Fast Algorithms for the Polynomial Time-Frequency Transform

This paper presents a general class of fast algorithms for computing the polynomial time-frequency transform (PTFT) of length-a^pb, where a, b, and p are positive integers. The process of derivation shows some interesting properties that are effectively used for minimization of the computational complexity. By assigning values of a, b, and p, various algorithms, for example, radix-o and split-radix-2/(2a), can be easily obtained to provide the flexibility supporting polynomial time-frequency transforms of various sequence lengths. The detailed analysis on the computational complexities needed by these algorithms is also presented in terms of the numbers of additions and multiplications. It is shown that the proposed algorithms significantly reduce the computational complexity for applications that deal with polynomial phase signals.