| Abstract: | Recently, joint spatial and spatial-frequency representations have been used in signal processing of non-stationary signals due to their natural local property and high joint resolution in both the spatial and spatial-frequency domain. However, a major obstacle to their implementation is their large computation requirements. This paper presents a fast n-dimensional Gabor transform and signal reconstruction algorithm employing multi-level parallel decomposition and fast Fourier transform techniques. The algorithm structure lends itself to implementation using VLSI/ASIC technology. Examples of two-dimensional Gabor transform and reconstruction performed on a AT computer demonstrate the substantial computational saving that can be achieved using the fast Gabor transform. |
