Rewind: A Music Transcription Method

Loading...
Thumbnail Image

Authors

Carthen, Chase D.

Issue Date

2016

Type

Thesis

Language

Keywords

Automatic Music Transcription , Deep Learning

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

Music is commonly recorded, played, and shared through digital audio formatssuch as wav, mp3, and various others. These formats are easy to use, but they lackthe symbolic information that musicians, bands, and other artists need to retrieveimportant information out of a given piece. There have been recent advances in theMusic Information Retrieval (MIR) field for converting from a digital audio format toa symbolic format. This problem is called Music Transcription and the systems builtto solve this problem are called Automatic Music Transcription (AMT) systems. Therecent advances in the MIR field have yielded more accurate algorithms using differenttypes of neural networks from deep learning and iterative approaches. Rewind’sapproach is similar but boasts a new method using an encoder-decoder network wherethe encoder and decoder both consist of a gated recurrent unit and a linear layer.The encoder layer of Rewind is a single layer autoencoder that captures the temporaldependencies of a song and produces a temporal encoding. In other words, Rewindis a web app that utilizes a deep learning method to allow users to transcribe, listento, and see their music.

Description

Citation

Publisher

License

In Copyright(All Rights Reserved)

Journal

Volume

Issue

PubMed ID

DOI

ISSN

EISSN