This repository holds 5 distinct datasets, each containing recordings of sounds of individual keystrokes. These datasets were created for classification purposes for the bachelor thesis The sound of typing: using Machine Learning to classify Keyboard Acoustic Emanations by Marcin Gólski, Piotr Kaszubski and Bartłomiej Woroch.
Each dataset contains 110 (100 train + 10 test) recordings of each of 43 physical
keys: A-Z, 0-9, '-' (dash), ';' (semicolon), "'" (quote), ','
(comma), '.' (period), '/' (forward slash) and ' ' (space).
Every train and test set is stored as a .wav file and a corresponding .keys file. A .keys file describes the offset in seconds of each keystroke within a corresponding .wav file, stored in plain text format.
For more information on how the datasets were recorded, their structure, and how to use them, please read the original paper and check the official source code repository: https://github.com/RoyalDonkey/put-kbd-thesis.