ORTHOGONALIZATION OF CIRCULAR STATIONARY VECTOR SEQUENCES AND ITS APPLICATION TO THE GABOR DECOMPOSITION

Certain vector sequences in Hermitian or in Hilbert spaces, can be orthogonalized by a Fourier transform, In the finite-dimensional case, the discrete Fourier transform (DFT) accomplishes the orthogonalization. The property of a vector sequence which allows the orthogonalization of the sequence by the DFT, called circular stationarity (CS), is discussed in this paper. Applying the DFT to a given CS vector sequence results in an orthogonal vector sequence, which has the same span as the original one, In order to obtain coefficients of the decomposition of a vector upon a particular nonorthogonal CS vector sequence, the decomposition is first found upon the equivalent DFT-orthogonalized one and then the required coefficients are found through the DFT, It is shown that the sequence of discrete Gabor basis functions with periodic kernel and with a certain inner product on the space of N-periodic discrete functions, satisfies the CS condition, The theory of decomposition upon CS vector sequences is then applied to the Gabor basis functions to produce a fast algorithm for calculation of the Gabor coefficients.