Winnow

A lightweight single-dimension Kalman filter for real-time data smoothing. Useful for reducing noise and impulses in slow or noisy signals where windowed averaging isn’t practical.

What it does

Filters incoming data by tracking the previous state and estimating the next with tunable system noise parameters.

Parameters

• Process noise: How much the true value changes (smaller = smoother but slower response)
• Sensor noise: How noisy your sensor is (higher = more filtering)
• Error estimate: Initial uncertainty
• Initial value: Starting point for the filter

Usage

C++ Version (winnow.hpp)

#include "winnow.hpp"

using winnow::Winnow;

Winnow filter(0.1, 0.5, 1.0, 0.0); // process noise, sensor noise, error estimate, initial value

double noisy_reading = readAnalogPin(analog_pin);
double clean_value = filter.getFilteredValue(noisy_reading);

C Version (Winnow.h)

#include "Winnow.h"

winnow_filter filter;
winnow_init(&filter, 0.1, 0.5, 1.0, 0.0); // process noise, sensor noise, error estimate, initial value

double noisy_reading = readAnalogPin(analog_pin);
double clean_value = winnow_get_filtered_value(&filter, noisy_reading);

License

MIT