Audio-Sense: Audio Fingerprint based Song and TV Channel Identification System

Abstract

The main objective of this project is to develop an audio fingerprint based song and TV channel identification system with use of Android mobile app for submitting the audio query. The system is called Audio-Sense. The well-known Shazam's landmark based audio fingerprint algorithm is used for representing the song and audio stream of the TV channels. The implementation of the Audio-Sense is based Dejavu open source project. For song identification, the fingerprints of these songs are first extracted off-line and stored in MySQL database. For TV channel identification, the fingerprints of the TV channel's audio are extracted in real-time and directly stored in the MySQL database. To tackle the problem of the outdated fingerprints, the TV audio fingerprints are periodically removed from the database. The developed Audio-Sense system supports web and mobile app queries. For web-based query, user can upload a MP3 audio clip to the server for searching the song database. For mobile app based query, user can use our developed Android audio sense app to record the audio in real-time and it will automatically upload to the server for song and TV channel identification.