Why SNR value is negative in my code ? Is the MSE value is correct?

DATASET

%%

clc;

clear

close all;

load B02T.mat;

fs=250;

% t = 0.004:1/fs:1;

eeg1 = data{1,1}.X;

channel_1= eeg1(:,1);

ch_1=channel_1(1:2000);

figure;

subplot(2,1,1);

plot(ch_1);

xlabel('Time (s)');

ylabel('Amplitude');

legend('ch1');

title("Plot channel "+1+ " for Structure 1X1");

hold on;

grid on;

eeg_signal = eeg1(1:2000); % first 2000 samples

subplot(2,1,2);

plot(eeg_signal,'black');

hold on;

grid on;

legend('EEG Signal');

title('Raw EEG Signal');

%%

[IMF, residual, info] = emd(eeg_signal);

imf_count = max(info.NumIMF);

figure;

for i=1:imf_count

subplot(6,2,i)

plot(IMF(:,i))

title("IMF"+i);

end

subplot(6,2,imf_count+1)

plot(residual)

title("Residue");

%%

thresholds = thselect(IMF, 'rigrsure'); % determine threshold for each IMF using universal thresholding

for i = 1:size(IMF, 2)

IMF(:, i) = wthresh(IMF(:, i), 's', thresholds(i));

end

% Reconstruct denoised signal from thresholded IMFs

denoised_signal = sum(IMF, 2) + residual;

clean_EEG = denoised_signal(1:2000);

% Plot results

figure;

subplot(2,1,1);

plot(eeg_signal);

title('Original signal');

subplot(2,1,2);

plot(clean_EEG);

legend( 'Denoised signal');

title('Denoised signal');

%%

original_signal = eeg_signal;

reconstructed_signal =clean_EEG;

% Calculate SNR

noise = original_signal - reconstructed_signal;

SNR = 20 * log10(norm(original_signal) / norm(noise));

% Calculate MSE

% MSE = mean((norm(original_signal) - norm(reconstructed_signal)).^2);

MSE = immse(norm(original_signal), norm(reconstructed_signal));

disp(['SNR: ', num2str(SNR), ' dB']);

disp(['MSE: ', num2str(MSE)]);

Answer 1

[Diwakar Diwakar]

In your code, the SNR value is negative because the calculation is incorrect. You are calculating the SNR using the formula:

SNR = 20 * log10(norm(original_signal) / norm(noise));

However, the denominator should be the norm of the noise, not just the noise itself. The correct calculation should be:

SNR = 20 * log10(norm(original_signal) / norm(noise));

Regarding the MSE value, you are using the `immse` function, which calculates the mean squared error between two signals. However, you are passing the norm of the signals instead of the signals themselves. To calculate the MSE correctly, you should use the original signals, like this:

MSE = immse(original_signal, reconstructed_signal);

[Komal]

What is the right formula and How to calculate SNR for geting right positive answer?