Você está na página 1de 16

Adaptive Line Enhancer (ALE) using Fast EDS

Presented by: Sarita Mishra-11528017 Vivek Surana -11550017

Project 12 : Adaptive Line Enhancer (ALE) using Fast EDS Generate a signal of the form where Ts is the sampling frequency and the frequencies f1 and f2 can be chosen as desired. Choose Ts so that the sinusoid appears smooth. 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 be about -7 dB. The desired signal (or primary signal) is d(n). The reference signal is d(n-1) 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 Fast EDS algorithm for adaptation. Experiment with different values of N and the forgetting factor . submit the following for your best case. (a) Plot the MSE for an ensemble of 100 runs. (b) Plot a few cycles of s(n) for one run. (c) Plot a few cycles of y(n) after convergence (one run). (d) Plot the impulse response and the frequency response of the filter after convergence (one run). I) (e) Find the SNR of the filtered signal. Make sure you correct for phase, if necessary. (f) Discuss the effects of N and on the performance of the algorithm. Provide all block diagrams, equations, and algorithms in your report. Discuss your results. Are they as expected? Are there any variations from your expectations? Comment on the convergence rate in your experiments. (g) Discuss the rate of convergence of the Fast EDS in relation to the EDS algorithm.

What are Adaptive Filter?


Filters with changing coefficients are called Adaptive filters .
Adaptive filters consists of the following two basic elements: A digital filter, which produces an output in response to an input

signal.
An adaptive algorithm, which adjusts the coefficients of the digital filter.

The effects of on the performance of the algorithm:


If =1 all the data are weighted equally and the algorithm has infinite memory length, which is optimal with respect to suppressing the estimation noise effect alone. If is small the algorithm has shorter memory length. Or Larger will reduce the estimation noise and small value reduces the equalization error due to lag effect. The optimal value of depends on channel fading dynamics and the extent of input noise on the equalization error.

The effects of N on the performance of the algorithm:


FEDS has O(N) computational complexity (4N+2). In this algorithm only
one directional search is performed for each sample of data so it has slow rate of convergence and as we increase the no of N the computational

complexity will increase and rate of convergence will decrease.


Rate of convergence of EDS is very much higher than FEDS. Because in EDS, N directions are searched for every sample.

Applications
System identification Sinusoidal tracking Noise cancellation Channel equalizer

Advantage: It reduces the computational complexity i.e. O(N) and required 4N+2 multiplications. Disadvantage: Its convergence rate is very low in comparison to EDS and RLS.

Adaptive line enhancher

Input signal
f1=input('enter the frequency of first signal') f2=input('enter the frequency of second signal') Ts1=1/(10*f1); Ts2=1/(10*f2); %%input signal... for i=1:1000

a(i)=5*sin(2*pi*f1*Ts1*i); b(i)=6*sin(2*pi*f2*Ts2*i); end s=a+b; Figure; plot(s) title('input signal') %%desired signal... ln=length(s); w=(10*rand(ln,1)-.5)';

Input signal (cont.)


Figure; plot(w) title('uniform white noise with 0 mean') Figure; d=s+w;; plot(d) title('desired signal d=s+w')

%%calculation of SNR p1=sum(s.^2); p1db=10*log10(p1); p2=sum(w.^2); p2db=10*log10(p2) SNR=p1db-p2db

OUTPUT
p1db = 33.0103

p2db = 40.7973

SNR = -7.7870

Você também pode gostar