• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

Automated Music Video Editing and Ranking

Student: Shvedov Denis

Supervisor: Ilya Makarov

Faculty: Faculty of Computer Science

Educational Programme: Financial Technology and Data Analysis (Master)

Year of Graduation: 2019

This is a first big part of a project called "Automated Music Video Editing and Ranking". The main task of this project is to create an automatic music video generation system that conducts soundtrack recommendation and video editing simultaneously. My colleague of this project is Pavel Adamenko, another HSE student. This paper is responsible for music recommendation part. It tries to compare all existing methodologies for classifying music by genre and mood and it presents our methodology for solution of this task. Relying on this classification, it will be much more easier to create music recommendation system. It will not just usual audio features, but audio features with some extra knowledge about music style and emotion that can really increase the quality of recommendation metrics. In the first chapter we attempt to describe music recommendation problem overall: why does good music recommendation system need for industry, why it is so hard to classify music by genre and mood. Also we briefly review the recent progress of the above problems with interesting solutions to solve it, including deep learning methods. In the second chapter we try to present our own methodology for classifying music by genre and mood. We deeply sink into feature extraction techniques and discuss the way with deep learning to solve the classification problems both. Here is some extra information about the architecture of convolution neural networks, which was used to classify music be genre and mood. In the final part of chapter the paper presents concept of music recommendation system, based on traditional recommendation system, and some metrics to understand the quality of this scheme. In the third chapter this paper presents experiments and results for genre/mood classification on chosen dataset and results for our own recommendation system. Also we compare values of our experiments to another method (with the same dataset). In the conclusion we try to summarize all the results in this work and to reason about future works. Particularly audio features from genre and mood classification will be used in the second part of project "Automated Music Video Editing and Ranking". In the bibliography everybody can find all the sources that was used in article.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses