Cover Song Identification Using Song-to-Song Cross-Similarity Matrix with Convolutional Neural Network

Lee, Juheon; Chang, Sungkyun; Choe, Sang Keun; Lee, Kyogu

doi:10.1109/icassp.2018.8461395

Cited by 9 publications

(4 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Secondly, a cross-similarity matrix of two tracks can be computed, and a convolutional network used to determine whether the inputs are versions of each other or not (i.e., binary classification) [46]. Although computing cross-similarity matrices introduces a computational load for the similarity estimation step, convolutional networks can replace the quadratic-complexity alignment algorithms like SWA, which results in a considerable improvement in terms of overall computational requirements.…”

Section: Classification-based Training -Classification-based Training...mentioning

confidence: 99%

Audio-based Musical Version Identification: Elements and Challenges

Yesiler,

Doras,

Bittner

et al. 2021

Preprint

View full text Add to dashboard Cite

Creating novel interpretations of existing musical compositions is and has always been an essential part of musical practice. Before the advent of recorded music, listening to a piece of music mostly meant listening to a version of it, in many cases, performed by musicians other than the original composer or performer. While this practice, along with musical notation, allows compositions to remain known for many decades or even centuries, it also provides room for artistic expression. Oftentimes, versions that are reproduced faithfully with respect to the original composition are seen as tributes to honor the composers; however, versions that are altered with the limitless creativity of humans often demonstrate how an existing idea can be transformed into something that goes beyond the original intention. Regardless of the ways in which musical versions are created, they are fundamental to the world's musical heritage.In this article, we consider a musical version to be "any rendition or performance of an existing musical composition." Another widely used term in the literature is "cover songs"; however, there are three main reasons we avoid using that term in this article. Firstly, the scope of the term is not clear as many authors limit its use to certain genres (e.g., pop and rock) and time periods (e.g., after the 1950s). Secondly, it has negative historical and economic connotations that date back to when the term emerged: record labels often asked White musicians to record (or cover) the songs of Black musicians as a marketing strategy to reach certain demographics. Lastly, the term does not represent the wide range of musical versions that are the subject of this article (see Fig. 1). Therefore, we employ the term "musical versions" throughout this article for being as inclusive as possible.This survey article focuses on the literature studying musical version identification and retrieval from a computational perspective. A key concept we refer to throughout this article is the concept of similarity, which we aim to quantify and model in a way that would reflect musical versions as semantically closer than non-versions. Following the literature, we consider that all versions of the same composition belong to the same group, or clique.

show abstract

Section: Classification-based Training -Classification-based Training...mentioning

confidence: 99%

Audio-based Musical Version Identification: Elements and Challenges

Yesiler,

Doras,

Bittner

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Copyright issues are analysed by investigating melody, the most defining, and the most predominant characteristic of any given piece of music. Similarly, melody tracking is also a primary tool in identifying cover songs (Juheon Lee, Sungkyun Chang, Donmoon Lee, 2015). Production tools are also enhanced by this feature, allowing musicians/producers to isolate melodies for later use, or remove a melody line from a given musical excerpt.…”

Section: Melody Trackingmentioning

confidence: 99%

High-Level Musical Content-Based Music Information Retrieval: A State-of-the-Art Review

Gedizlioglu¹,

Erol²

2020

JSTR

View full text Add to dashboard Cite

Music Information Retrieval (MIR) is an interdisciplinary field that involves automating music processing for the purpose of accessing and managing large music collections. Motivated by the rapid growth of digital music content, it encompasses a broad set of strategies from diverse disciplines. In this paper, we provide a task-centric perspective of this field with particular emphasis on high-level content analysis. We provide a general context to explain the contributions from various disciplines for those wishing to learn about this field. We also discuss challenges and future directions to stimulate further research.

show abstract

“…At the lower end of the specificity scale are tasks such as genre recognition [36]. A medium-level specificity is associated with tasks such as audio matching [12,14], version identification [31,37], live song detection [28,38], and cover song retrieval [8,10,13,15,16,[21][22][23][24][25][26][27]29,30]. In all of these tasks, one allows for variations as they typically occur in different performances and arrangements of a piece of music.…”

Section: Related Workmentioning

confidence: 99%

Learning Low-Dimensional Embeddings of Audio Shingles for Cross-Version Retrieval of Classical Music

Zalkow

Müller

2019

Applied Sciences

View full text Add to dashboard Cite

Cross-version music retrieval aims at identifying all versions of a given piece of music using a short query audio fragment. One previous approach, which is particularly suited for Western classical music, is based on a nearest neighbor search using short sequences of chroma features, also referred to as audio shingles. From the viewpoint of efficiency, indexing and dimensionality reduction are important aspects. In this paper, we extend previous work by adapting two embedding techniques; one is based on classical principle component analysis, and the other is based on neural networks with triplet loss. Furthermore, we report on systematically conducted experiments with Western classical music recordings and discuss the trade-off between retrieval quality and embedding dimensionality. As one main result, we show that, using neural networks, one can reduce the audio shingles from 240 to fewer than 8 dimensions with only a moderate loss in retrieval accuracy. In addition, we present extended experiments with databases of different sizes and different query lengths to test the scalability and generalizability of the dimensionality reduction methods. We also provide a more detailed view into the retrieval problem by analyzing the distances that appear in the nearest neighbor search.

show abstract

Cover Song Identification Using Song-to-Song Cross-Similarity Matrix with Convolutional Neural Network

Cited by 9 publications

References 5 publications

Audio-based Musical Version Identification: Elements and Challenges

Audio-based Musical Version Identification: Elements and Challenges

High-Level Musical Content-Based Music Information Retrieval: A State-of-the-Art Review

Learning Low-Dimensional Embeddings of Audio Shingles for Cross-Version Retrieval of Classical Music

Contact Info

Product

Resources

About