[go: up one dir, main page]
Scribd
Upload
45 views20 pages

Ale Using Flms Algorithm: Project On

This document summarizes a project on adaptive line enhancement (ALE) using the fast least mean square (FLMS) algorithm. It compares FLMS to block least mean square (BLMS) and conventional LMS. FLMS is more computationally efficient by implementing BLMS in the frequency domain. The document tests different block lengths, finding optimal convergence for a length matching the filter length. It also compares FLMS and LMS, finding FLMS converges faster. Finally, it explores using partially overlapping blocks for FLMS, finding increased overlap improves convergence speed at the cost of higher computation.

Uploaded by

Pinky Bhagwat
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
45 views20 pages

Ale Using Flms Algorithm: Project On

This document summarizes a project on adaptive line enhancement (ALE) using the fast least mean square (FLMS) algorithm. It compares FLMS to block least mean square (BLMS) and conventional LMS. FLMS is more computationally efficient by implementing BLMS in the frequency domain. The document tests different block lengths, finding optimal convergence for a length matching the filter length. It also compares FLMS and LMS, finding FLMS converges faster. Finally, it explores using partially overlapping blocks for FLMS, finding increased overlap improves convergence speed at the cost of higher computation.

Uploaded by

Pinky Bhagwat
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 20

Project on

ALE USING FLMS ALGORITHM


Prepared by : Mr Prabhash K Singh (Communication System,12531012)

Mr. Sreekanth CB (Communication System,12531014)

BLMS(Block Least Mean Square)


Incoming data sequence is sectioned into blocks using serial to parallel converter. Block of input data are applied to FIR filter one block at a time. Adaptation of filter proceeds on a block by block basis rather than sample by sample basis as in conventional LMS algorithm.

FLMS(Fast Least Mean Square)


BLMS requires both linear convolution and linear correlation in its implementation. FFT provides powerful tool for faster convolution and faster correlation. A frequency domain implementation of BLMS will be more efficient. FLMS is the frequency domain implementation of BLMS Computationally more efficient than BlockLMS(more than 15 times).

FLMS Algorithm

Algorithm continued..

ALE Using FLMS Algorithm


Generate a signal of the form s(n) = A sin(2f1n/Fs) + B sin(2f2n/Fs) Fs = sampling Frequency A noise w(n) is added to the signal to produce the noisy signal d(n) = s(n) + w(n) The noise is a white uniformly distributed sequence with zero mean. The SNR of this signal should about -7 dB. The desired signal ( or primary signal) is d(n). The reference signal is d(n1) and is applied to the input of an N-tap FIR filter that produces the output y(n), which should be the estimate of the signal. Use the LMS algorithm for adaptation. Choose N and the step size appropriately for good results.

Part one:
Experiment with different block lengths of L=100, 200, 400, 800, and 1000. Discuss the effect of the block length on convergence, if any. How does the computational change with the block length? Using SNR=10

Fig 1: Block length=100

Fig 2: Block length=200

Fig 3: Block length=400

Fig 4: Block length=800

Fig 5: Block length=1000

Effect of Block length on convergence: when block length increases 1. convergence didnt happen. To maintain convergence we have to decrease the step size. 2. computational complexity decreases for same number of iterations

Optimal choice is block length = filter length, its computationally more efficient. When block length < filter length -advantage of reduced processing delay -still more efficient than conventional LMS. block length > filter length there is redundant operations in adaptive process

Part 2:

FLMS v/s LMS

FLMS is computationally more efficient. FLMS converges at a faster rate than LMS. Convergence rate of FLMS can be further improved by assigning different step size for each weight.

Part 3:
In the FLMS, contiguous blocks of data are used. One variantion is to use partially overlapping blocks in order to gain convergent speed. Use overlaps of L/4, L/2 and 3L/4 and repeat your experiments Discuss the performance of the algorithm.

As overlapping between block elements increases their convergence is also faster. But need more computation is involved.

!!! THANK YOU !!!

You might also like