FPGA Implementation of MRMN with Step-Size Scaler Adaptive Filter for Impulsive Noise Reduction
- PDF / 2,762,765 Bytes
- 29 Pages / 439.37 x 666.142 pts Page_size
- 100 Downloads / 190 Views
FPGA Implementation of MRMN with Step-Size Scaler Adaptive Filter for Impulsive Noise Reduction Priyank H. Prajapati1 · Anand D. Darji1 Received: 5 April 2019 / Revised: 26 December 2019 / Accepted: 27 December 2019 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Noise reduction is an essential part of the signal processing for error-free analysis and critical measurements of the signal. Various robust mixed norm (RMN)-based adaptive algorithms have been reported to remove Gaussian and impulsive noise together. In this paper, a modified robust mixed norm (MRMN) with step-size scaler-based adaptive filter has been proposed to suppress the impulsive and Gaussian noise in system identification application. Further, an attempt has been made to develop an algorithm that simultaneously improves the rate of convergence and reduce the steadystate error (SSE). The proposed adaptive algorithm on an average decreases 13.2% SSE at the same initial rate of convergence, and at the same SSE, the rate of convergence is increased by 43.8% as compared to the existing mixed norm-based adaptive algorithms. Moreover, the hardware architecture of the proposed algorithm has been implemented using VHDL on various FPGA platforms. The proposed hardware implementation of the weight update block leads to high-speed realization of the adaptive filter with little increases in hardware resources. The architecture offers a maximum clock frequency of 66.53 MHz when implemented on Virtex 5 FPGA. Keywords Noise · Adaptive filter · MRMN · SSE · Step-size scaler · FPGA
1 Introduction The system design based on digital signal processing applications is highly growing in the various field of biomedical, communication, nuclear fusion, RADAR, and seismology engineering. The most often problem faced by system designers is inaccuracy due to the noise (an unwanted signal) present in the signal, which may lead to an undesired
B
Priyank H. Prajapati [email protected] Anand D. Darji [email protected]
1
Sardar Vallabhbhai National Institute of Technology, Ichchhanath, Surat, Gujarat, India
Circuits, Systems, and Signal Processing
system output [6,34]. There is mainly Gaussian and impulsive kind of noise present in the signal, which distorts the signals severely, and it may create serious measurement errors [27]. It is a motivational and essential part of signal processing to remove these noises in order to improve the system performance or reduce the measurement errors. 1.1 Adaptive Filter for Noise Reduction In literature [1,2,4,8,9,12,16,17,22,23,25–28,31,35–38,41], different algorithms have been reported to remove the impulsive and Gaussian noise. The clipping-based methods [9,22,35,38] remove the impulsive noise to some extent, but they are not accurate because, in this method, the position of impulse noise must be detected first and then corrected. On the other hand, the adaptive filter can detect the impulse noise irrespective of the impulse position in the signal and corrects them [1,2,4,8,12,16,17,
Data Loading...
