No Detail Found!

Performing lowpass, highpass, bandpass and bandstop filtering of input auto and cross correlation sequences in linear time.

Author(s)
Ejaz A. Ansari Department of Electrical Engineering, COMSATS Institute of Information Technology, Lahore, Pakistan
Abstract
Performing filtering on the input and output data sequences through linear time invariant (LTI) systems formulates an important operation of Digital Signal and Image Processing fields. This paper proposes a simple and efficient technique in discrete-time domain which does the lowpass, highpass, bandpass and bandstop filtering of the input auto and cross correlation sequences of discrete-time real wide sense stationary (WSS) process through exponential, rectangular, triangular and trapezoidal shaped windows in linear time. As the frequency responses of the discrete-time windows mentioned above resemble very much with the frequency response of low pass filters, the proposed algorithm [5] does indeed perform the Low Pass Filtering (LPF) of the input auto and cross correlation sequences in linear time. We have already shown in [5] that processing or LPF of auto and cross correlation sequences through the above four mentioned windows can be achieved in O(L) time, where L denote the size of the input autocorrelation sequence. From the knowledge of impulse response of low pass filter, hlp[n], the impulse response of a high pass filter hhp[n] may be obtained by subtracting hlp[n] from the unit sample sequence, δ[n]. This shows that high pass filtering (HPF) of the input auto and cross correlation sequences can also be performed through above four windows in linear time. Finally, the known values of impulse responses of low and high pass filters determine the impulse and frequency responses of a band pass filter, hbp[n]. In frequency domain, it is usually implemented with a cascade connection of low and high pass filter’s frequency responses with different cut-off frequencies. However, in time domain, hbp[n] becomes equal to the convolution of hlp[n] with hhp[n]. Moreover, like hhp[n], the impulse response of band stop filter, hbs[n] is simply equal to the difference of δ[n] and hbp[n]. This reveals that bandpass filtering (BPF) and bandstop filtering (BSF) operations on the input auto and cross correlation sequences can also be done through above four windows in linear time. We have also shown in [5] that our proposed algorithm outperformed all the existing techniques in both time and frequency domains with regards to both exact number of arithmetic operations and to its worst case time complexity that grew linearly with the length of the input auto correlation sequence for the case when L >> K, where K denote the size of an exponential window.
Publication Details
Page(s) 621-626
DOI DOI not available
Published Journal: Science International, Volume: 26, Issue: 2, Year: 2014

Request the full text from PASTIC

0

Citations

0

Downloads

63

Views