Platforma za klasifikacijo zvočnih posnetkov

BUDNA, BORUT (author), Curk, Tomaž (mentor), Gams, Matjaž (co-mentor)

V diplomskem delu opišemo sistem za obdelavo in klasifikacijo zvočnih posnetkov, ki smo ga zasnovali in razvili kot dodatek za program Orange. Orange je odprtokodni sistem, namenjen podatkovnemu rudarjenju. Vsebuje gradnike za prikazovanje in manipuliranje s podatki, gradnjo napovednih modelov, in gradnike za mnoga druga opravila. Ne vsebuje gradnikov za branje in obdelovanje zvočnih zapisov. Zato smo v ta namen izdelali dodatek, imenovan ``Audio - IJS,'' znotraj katerega smo razvili osem gradnikov, ki omogočajo manipulacijo z zvokom in integracijo z ostalimi gradniki v Orange. Izdelani gradniki omogočajo filtriranje, segmentacijo (razrez), spektralno analizo, ekstrakcijo značilk, predvajanje zvoka in še nekaj ostalih funkcij, ki so potrebne za povezovanje s preostalimi gradniki Orange. Razviti dodatek uporabniku omogoča hitro izdelavo sofisticiranih analiz zvočnih posnetkov.

• Language: Slovenian
• Keywords: Orange, podatkovno rudarjenje, filtriranje, segmentacija, ekstrakcija značilk, spektrogram, klasifikacija
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2017
• PID: 20.500.12556/RUL-95070
• Publication date in RUL: 13.09.2017

Secondary language

• Language: English
• Title: Platform for audio clips classification

Abstract

In this diploma thesis we will describe audio clip analysis and classification system that we designed and developed as an add-on for Orange software. Orange is an open-source system, designed for data mining. It contains widgets for displaying and manipulating data, building predictive models and several other tasks. It does not, however, contain widgets for reading and processing audio clips. For this purpose, we built an add-on called »Audio-IJS« within which we developed eight widgets that enable audio manipulation as well as integration with other Orange widgets. The constructed widgets enable filtration, segmentation (cutting), spectral analysis, feature extraction, audio play as well as several other functions, necessary for connecting with other Orange widgets. The developed add-on enables the user a rapid creation of sophisticated analyses of audio clips.

• Keywords: Orange, data mining, filtering, segmentation, feature extraction, spectrogram, classification