This project was inspired by the popular "secret sound" radio contest. Stations would play a sound clip (usually ~5 sec) and contestants would dial into guess what the sound was. Sometimes, it could take more than ten guesses before any contestent got the right answer. I decided to create a generic sound recognizer algorithm as a hack for the secret sound contest. Such an algorithm, however, has potential beyond radio show contests (lol), and can hopefully be used to build smarter sensors and better AI.
This project consists of two pieces: a scraper to pull audio files from the web, and a recognizer that is trained by these audio clips and attempts to recognize other sounds. For more information on the scraper, please visit the folder:
scraper
For more information on the recognizer, please visit the folder:
recognizer