Identifikacija glasbe v videoposnetkih s pomočjo zvočnih prstnih odtisov

REPŠE, BINE (author), Marolt, Matija (mentor)

Diplomsko delo opisuje postopek izdelave aplikacije, ki analizira videoposnetke ter s pomočjo zvočnih prstnih odtisov prikaže vse pesmi, ki jih je v posnetku moč slišati, poleg tega pa prikaže, kdaj se posamezna pesem v posnetku prične in kdaj se konča. Aplikacija uporabniku omogoča tudi uporabo zunanje spletne storitve za odkrivanje pesmi, katerih zvočnih prstnih odtisov uporabnik nima shranjenih v podatkovni bazi. Na to storitev se mora uporabnik naročiti. V primeru, ko pa skladbe niso prepoznane, aplikacija uporabniku nudi možnost vnosa podatkov o teh neprepoznanih skladbah ter shranjevanju le teh v podatkovno bazo za zmožnost prepoznavanja v prihodnosti. V nalogi opišem obstoječe rešitve za prepoznavanje glasbe ter podrobneje opišem delovanje knjižnice, ki sem jo za ta namen izbral. Opišem tudi celoten postopek implementacije, za zaključek pa prikažem še rezultate testiranja aplikacije ter predlagam možne izboljšave za prihodnost.

• Language: Slovenian
• Keywords: zvočni prstni odtisi, videoposnetki, prepoznavanje
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2016
• PID: 20.500.12556/RUL-85132
• Publication date in RUL: 13.09.2016

Secondary language

• Language: English
• Title: Audio fingerprinting for music identification in videos

Abstract

The thesis describes the development process of making an application, the purpose of which is to analyze video files and to recognize the songs that appear in the videos, using audio fingerprints. The application delivers the information about where within the duration of a video a certain song begins and where it ends. The application also enables its user the use of an online service used to recognize songs which could not have been recognized using the original, free library. The user has to subscribe to this online service and pay for its use. In case the application is not able to recognize the songs within the recording it allows the user to write the information about the unrecognized songs and to fingerprint them and save them into the database.The thesis describes the solutions for recognizing music and provides a detailed description of getting from an audio file to a set of audio fingerprints using the library of my choice. It also describes the process of making the application and finally, presents some testing results and proposes some possible improvements.

• Keywords: audio fingerprints, video, recognition