Filter Peaks Below a Threshold with findpeaks in MATLAB

To exclude unwanted noise or small peaks in MATLAB, use the 'MinPeakHeight' or 'MinPeakProminence' name-value pairs in the findpeaks() function.

1. Method 1: Absolute Amplitude Threshold ('MinPeakHeight')

This filters out any peak whose absolute value falls below a specified minimum level (e.g., threshold = 1.5).

% Generate sample noisy signal with peaks of different heights
t = linspace(0, 10, 1000);
signal = sin(t) + 0.8*cos(3*t) + 0.5*sin(7*t) + 0.1*randn(size(t));

% Define threshold
minHeight = 1.2;
% Find peaks strictly greater than minHeight
[pks, locs] = findpeaks(signal, t, 'MinPeakHeight', minHeight);
% Plot signal and detected peaks
figure('Color', 'w', 'Position', [100, 100, 750, 400]);
plot(t, signal, 'b-', 'LineWidth', 1.2, 'DisplayName', 'Signal');
hold on;
plot(locs, pks, 'ro', 'MarkerFaceColor', 'r', 'MarkerSize', 6, ...
    'DisplayName', sprintf('Peaks > %.1f', minHeight));
yline(minHeight, 'k--', 'LineWidth', 1.5, ...
    'DisplayName', sprintf('Threshold (%.1f)', minHeight));
xlabel('Time (s)');
ylabel('Amplitude');
title('Peak Filtering Using MinPeakHeight');
legend('Location', 'northeast');
grid on;

2. Method 2: Prominence-Based Filtering ('MinPeakProminence')

When the signal baseline drifts up and down, absolute height thresholds fail. Use 'MinPeakProminence' to measure how much a peak stands out relative to its local surrounding valleys.

% Signal on a drifting baseline
driftingBaseline = 0.5 * t;
noisySignal = sin(2*pi*0.5*t) + driftingBaseline + 0.15*randn(size(t));

% Detect only significant peaks regardless of baseline height
minProminence = 1.0;
[pks_prom, locs_prom] = findpeaks(noisySignal, t, ...
    'MinPeakProminence', minProminence);
figure('Color', 'w');
plot(t, noisySignal, 'Color', [0.3 0.3 0.8], 'DisplayName', 'Drifting Signal');
hold on;
plot(locs_prom, pks_prom, 'rv', 'MarkerFaceColor', 'r', 'MarkerSize', 7, ...
    'DisplayName', 'Significant Peaks');
xlabel('Time (s)');
ylabel('Amplitude');
title('Prominence-Based Peak Detection (Handles Baseline Drift)');
legend('Location', 'northwest');
grid on;

3. Combining Multiple Peak Constraints

Combine height, prominence, and minimum distance between peaks for robust noise rejection.

[pks, locs, w, p] = findpeaks(signal, t, ...
    'MinPeakHeight', 1.0, ...      % Absolute height threshold
    'MinPeakProminence', 0.5, ...  % Prominence threshold
    'MinPeakDistance', 0.8);       % Minimum time spacing between peaks

% Built-in plot visualization
figure('Color', 'w');
findpeaks(signal, t, ...
    'MinPeakHeight', 1.0, ...
    'MinPeakProminence', 0.5, ...
    'MinPeakDistance', 0.8);
grid on;

Parameter Selection Guide