Frequency warping is theoretically designed to be a unitary operator of infinite input and output dimensions, thus performing the resolution of identity. In real implementations finite dimensions have to be considered, then perfect reconstruction cannot be fulfilled. The accuracy of reconstruction is particularly compromised in case of non-smooth warping maps, which are more useful for practical applications. In order to overcome this limitation, a new frequency warping biorthogonal frame operator for non-smooth warping maps is introduced in this work. The proposed transformation is based on a mathematical model which has been previously introduced for computational purposes. By adding some redundancy with respect to the truncation of the infinite dimensions operator, the effect of an infinite output dimension can be taken into account in a compressed way, based on an analytical factorization. In the reconstruction process, the additional redundant samples are expanded, thus guaranteeing near perfect reconstruction.
S. Caporale, L. De Marchi, N. Speciale (2011). Frequency Warping Biorthogonal Frames. IEEE TRANSACTIONS ON SIGNAL PROCESSING, 59, 2575-2584 [10.1109/TSP.2011.2114658].
Frequency Warping Biorthogonal Frames
CAPORALE, SALVATORE;DE MARCHI, LUCA;SPECIALE, NICOLO'ATTILIO
2011
Abstract
Frequency warping is theoretically designed to be a unitary operator of infinite input and output dimensions, thus performing the resolution of identity. In real implementations finite dimensions have to be considered, then perfect reconstruction cannot be fulfilled. The accuracy of reconstruction is particularly compromised in case of non-smooth warping maps, which are more useful for practical applications. In order to overcome this limitation, a new frequency warping biorthogonal frame operator for non-smooth warping maps is introduced in this work. The proposed transformation is based on a mathematical model which has been previously introduced for computational purposes. By adding some redundancy with respect to the truncation of the infinite dimensions operator, the effect of an infinite output dimension can be taken into account in a compressed way, based on an analytical factorization. In the reconstruction process, the additional redundant samples are expanded, thus guaranteeing near perfect reconstruction.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.