Implementation of an audio fingerprinting technique. Evaluation of its accuracy and robustness

González Toni, Carla M. (2017). Implementation of an audio fingerprinting technique. Evaluation of its accuracy and robustness. Proyecto Fin de Carrera / Trabajo Fin de Grado, E.T.S.I. y Sistemas de Telecomunicación (UPM), Madrid.

Description

Title: Implementation of an audio fingerprinting technique. Evaluation of its accuracy and robustness
Author/s:
  • González Toni, Carla M.
Contributor/s:
  • Virtanen, Tuomas
Item Type: Final Project
Degree: Grado en Ingeniería de Sonido e Imagen
Date: 27 July 2017
Subjects:
Freetext Keywords: Huella de audio
Faculty: E.T.S.I. y Sistemas de Telecomunicación (UPM)
Department: Otro
Creative Commons Licenses: Recognition - No derivative works - Non commercial

Full text

[img]
Preview
PDF - Requires a PDF viewer, such as GSview, Xpdf or Adobe Acrobat Reader
Download (1MB) | Preview

Abstract

El objetivo principal del proyecto es implementar una técnica de audio fingerprinting para lo que se desarrollará un programa en MATLAB que permita identificar una señal de audio dentro de una base de datos. Así mismo se evaluará su eficacia y su robustez. El concepto de audio fingerprinting hace referencia a un sistema identificador de señales de audio basado en el contenido del archivo. Con esta "huella de audio" se puede identificar un patrón, para luego poder reconocer el audio sin la necesidad de tener ningún tipo de información sobre este. El proyecto se dividirá principalmente en tres partes. Para empezar, será necesario crear una base de datos lo suficientemente amplia (alrededor de mil pistas). Una vez que la base está creada, se introducirá una pista al programa y éste tendrá que comprobar si se encuentra en la base de datos o no. Si la pista de audio sí que está dentro, se devolverá el nombre de la pista encontrada así como el porcentaje de semejanza. La tercera parte constará en evaluar la robustez del programa introduciendo señales de audio distorsionadas con distintos tipos de ruido (blanco, de coches y de murmullos) y con distintos niveles de ruido. Si se considera que el diseño no es suficientemente eficaz, se procederá a mejorar el diseño del algoritmo elegido previamente. Abstract: The main goal of this thesis is to implement an audio fingerprinting technique. It will be developed a program coded in MATLAB that allows to identify an audio signal inside a database. Besides it will be evaluated its accuracy and robustness. The concept of audio fingerprinting refers to an identifier system of audio signals based on the archive content. With this audio fingerprinting it is possible to identify a pattern, for later being able to recognize the audio without the necessity of having any kind of information about it. The process will be divided mainly in three parts. Firstly, it is necessary to create a wide data set (around one thousand tracks). Once this data set is created, one track is introduced to the program and the program has to check if the track is already belongs to the database or not. If the track is inside, the program will return the name of the introduced track as well as the matching percentage of similarity. The third part consists on evaluate the accuracy and the robustness of the program introducing audio signals distorted with different kinds of noise (white noise, car noise and babble noise) and with different levels of it. If after this evaluation is considered that the design is not enough effective, it will proceed to improve the design of the algorithm chosen earlier.

More information

Item ID: 52610
DC Identifier: http://oa.upm.es/52610/
OAI Identifier: oai:oa.upm.es:52610
Deposited by: Biblioteca Universitaria Campus Sur
Deposited on: 16 Oct 2018 10:30
Last Modified: 16 Oct 2018 10:30
  • Logo InvestigaM (UPM)
  • Logo GEOUP4
  • Logo Open Access
  • Open Access
  • Logo Sherpa/Romeo
    Check whether the anglo-saxon journal in which you have published an article allows you to also publish it under open access.
  • Logo Dulcinea
    Check whether the spanish journal in which you have published an article allows you to also publish it under open access.
  • Logo de Recolecta
  • Logo del Observatorio I+D+i UPM
  • Logo de OpenCourseWare UPM