Optimal FPGA implementation of CL multiwavelets architecture for signal denoising application

Author: Kumar B. Mohan   Lavanya R. Vidhya   Sumesh E.P.  

Publisher: Taylor & Francis Ltd

ISSN: 1362-3060

Source: International Journal of Electronics, Vol.100, Iss.3, 2013-03, pp. : 288-301

Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.

Previous Menu Next

Abstract

Wavelet transform is considered one of the efficient transforms of this decade for real time signal processing. Due to implementation constraints scalar wavelets do not possess the properties such as compact support, regularity, orthogonality and symmetry, which are desirable qualities to provide a good signal to noise ratio (SNR) in case of signal denoising. This leads to the evolution of the new dimension of wavelet called ‘multiwavelets’, which possess more than one scaling and wavelet filters. The architecture implementation of multiwavelets is an emerging area of research. In real time, the signals are in scalar form, which demands the processing architecture to be scalar. But the conventional Donovan Geronimo Hardin Massopust (DGHM) and Chui-Lian (CL) multiwavelets are vectored and are also unbalanced. In this article, the vectored multiwavelet transforms are converted into a scalar form and its architecture is implemented in FPGA (Field Programmable Gate Array) for signal denoising application. The architecture is compared with DGHM multiwavelets architecture in terms of several objective and performance measures. The CL multiwavelets architecture is further optimised for best performance by using DSP48Es. The results show that CL multiwavelet architecture is suited better for the signal denoising application.