Intercoder-Reliability (ICR) is a small experiment for testing whether information can survive a signal-processing pipeline without being corrupted beyond recognition.
The project converts text into binary-like signal data, applies filters such as Gaussian and Butterworth filters, reconstructs the data back into a string, and compares it with the original. The goal is to estimate whether the transmission path preserves the message well enough to recover the original contents.
The main concept is:
- Convert a string such as
Hello Worldinto binary data. - Represent the binary data as a NumPy array (
ndarray). - Process the array through a signal filter.
- Convert the filtered data back to binary and then back to text.
- Compare the reconstructed string with the original string.
- Decide whether the transmission was successful.
This is a simple analog-to-digital/filtered-signal demonstration inspired by communication-system reliability tests.
- Hello-World-Reliability.ipynb — basic reliability experiment for a text message.
- Binary-reliability1.ipynb — additional binary-oriented reliability exploration.
- bmp-reliability.ipynb — reliability experiment using bitmap-style data.
- bmp-colortomonochrome.ipynb — conversion from color BMP data to monochrome output.
The experiments use SciPy filtering operations:
- Gaussian filter:
scipy.signal.windows.gaussianwith convolution - Butterworth filter:
scipy.signal.butterwithlfilter
These filters are used to simulate a noisy or constrained transmission channel and test how well the original message survives processing.
This project uses Python with dependencies managed through uv.
uv syncIf you want to start Jupyter locally:
uv run jupyter notebookor:
uv run jupyter labOpen the notebook files in Jupyter and execute the cells in order. Each notebook demonstrates a different variant of the same idea: encode information, filter it, decode it, and compare the result.
A successful run should preserve the original text closely enough that the recovered message matches the input text (or differs only in a limited, explainable way). The project is intended to visualize how filtering affects signal fidelity and whether the message is still recoverable.
Gaussian filter example:
Butterworth filter example:
This repository is best understood as a lightweight reliability testbed for signal-processing ideas rather than as a packaged library. It is most useful when used interactively through the notebooks.



