A hybrid recurrent neural network for music transcription

Siddharth Sigtia; Emmanouil Benetos; Nicolas Boulanger-Lewandowski; Tillman Weyde; Artur S. d'Avila Garcez; Simon Dixon

ICASSP 2015accepted0 citations

A hybrid recurrent neural network for music transcription

Siddharth Sigtia, Emmanouil Benetos, Nicolas Boulanger-Lewandowski, Tillman Weyde, Artur S. d'Avila Garcez, Simon Dixon

Abstract

We investigate the problem of incorporating higher-level symbolic score-like information into Automatic Music Transcription (AMT) systems to improve their performance. We use recurrent neural networks (RNNs) and their variants as music language models (MLMs) and present a generative architecture for combining these models with predictions from a frame level acoustic classifier. We also compare different neural network architectures for acoustic modeling. The proposed model computes a distribution over possible output sequences given the acoustic input signal and we present an algorithm for performing a global search for good candidate transcriptions. The performance of the proposed model is evaluated on piano music from the MAPS dataset and we observe that the proposed model consistently outperforms existing transcription methods.

BibTeX

@inproceedings{icassp2015_ahybridrecurrent,
  title = {A hybrid recurrent neural network for music transcription},
  author = {Siddharth Sigtia and Emmanouil Benetos and Nicolas Boulanger-Lewandowski and Tillman Weyde and Artur S. d'Avila Garcez and Simon Dixon},
  booktitle = {ICASSP 2015},
  year = {2015}
}