FIR adaptive filter implementation using LMS/RLS algorithm with DSP system toolbox

Hi hello,

I am trying to implement FIR adaptive filter with LMS/RLS algorithm for noise reduction in my acquired ppg signal. The ppg signal is primary input a+b where a is the desired signal not correlated with r (noise reference input of the filter) b is the noise of the signal ppg correlated with r. I am using dsp.LMSFilter, but I am getting exactly same output as the given input signal. I am not sure, if I have used the DSP toolbox functions correctly as well, for which I have attached my code below. I have followed exactly - https://de.mathworks.com/help/dsp/ug/enhance-a-signal-using-lms-and-normalized-lms-algorithms.html

% Read the data
data = readmatrix('U:\1_thesis\work_package_3\log_data_analysis\filtered.csv');
% x - desired ppg signal and v2 - noise ref acc data
x = data(:,1);
v2 = data(:,2);

% setting up the filter
L = 2;
lms = dsp.LMSFilter(L,'Method','LMS');
[mumaxlms,mumaxmselms] = maxstep(lms,x);
lms.StepSize  = mumaxmselms/30;

% y1 - filtered signal, elms - error
[y1,elms,wlms] = lms(v2,x);

% total data length is 15002
n = (1:15002)';

% output graph
plot(n(1:1200),[y1(1:1200),elms(1:1200)]);
legend('LMS output signal', 'LMS error');
xlabel('Time index (n)');
ylabel('Amplitude');
hold on;
plot(n(1:1200),x(1:1200),'k:'); %input noisy signal plot
xlabel('Time index (n)');
ylabel('Amplitude');
hold off;

Your help is very much appreciated and Thanks in advance

[Pratheek]

Please provide the filtered.csv file to help you better.

Answer 1

[Balaji]

Hi Siddarth,

As per my understanding, you have issue while obtaining desired results using “dsp.LMSFilter”. You have correctly implemented the filter, but in order to get better results, decrease the “lms.StepSize” for a small steady state error between the output and the desired signal.

You can refer to the documentation of “dsp.LMSFilter” below:

https://in.mathworks.com/help/dsp/ref/dsp.lmsfilter-system-object.html