Design and Implementation of an Audio Fingerprinting System for the Identification of Audio Recordings

Rapid growth in the multimedia industry due to various streaming platforms has increased audio and video traffic enormously. This traffic generates a massive amount of data that requires efficient algorithms to retrieve the desired data in a short amount of time. Thus, there is a need for efficient audio retrieval methods such as audio fingerprinting. Audio fingerprinting systems mainly use audio signals after processing to obtain a representative hash called an audio fingerprint. The fingerprint holds content information of a recording that can distinguish one recording from another by extracting relevant features from the audio content. Some applications of audio fingerprinting include Music Retrieval, Copyright infringement, digital watermarking and broadcast monitoring. We propose to build a reliable audio fingerprinting system, which uses a robust audio fingerprint extraction method and an efficient search strategy, which uses only limited computing resources, with minimized search time for recognition of audio content, by considering other musical features. Accuracy, confidence, efficiency in storing the fingerprints and speed is used to measure the system's performance. The accuracy of the system depends on the confidence level of the match found. The accuracy varies with the recording time. The ideal recording time is found to be 5 seconds that recognizes the song with accuracy of 83.3%. The system performs well even in the presence of noise with reduced false positives.