Efficient VLSI architectures for the biorthogonal wavelet transform by filter bank and lifting scheme

Design Optimization of Integer Lifting DWT Circuitry for Implantable Neuroprosthetics

“…Thus, by processing three time samples simultaneously, it is possible to execute the filtering steps in real time. It has been shown that a pipeline structure is very efficient at executing the steps of a liftingbased wavelet transform [5][6]. Fig.…”

“…implementations, a detailed study was performed to determine the hardware requirements for expanding the circuits to an arbitrary number of channels and levels. The models developed to describe the area, A, and power consumption, P, of the multi-channel/level implementations are given by: with channel and level, respectively; C is the number of channels; L is the number decomposition levels; Cf and Lf, are the clocking frequency scaling factors for channel and level, respectively; and Lu, is a level usage factor, which is related to the general level usage factor U(L) given by 2L l (6) Table II lists a simplified set of the coefficients used to model both configurations. Data generated by these models, including plots of the area-power product versus channel and level shown in Fig.…”

“…Recent results indicate that signal integrity can be preserved using integer wavelet transform (IWT) with quantized filter coefficients [2]. Unfortunately, most of the reported hardware approaches for lifting-based wavelet transform computation [3][4][5][6] focus on speed performance and do not adequately address power and area constraints. Moreover, it is assumed that an entire data frame is available, making them unsuitable for streaming neural data in real-time.…”