Amplitude Spectrogram Prediction from Mel-Frequency Cepstrum Coefficients Using Deep Neural Networks

Timbre conversion of musical instrument sounds, utilizing deep neural networks (DNNs), has been extensively researched and continues to generate significant interest in the development of more advanced techniques. We propose a novel algorithm for timbre conversion that utilizes a variational autoenc...

Full description

Saved in:
Bibliographic Details
Published inJournal of Signal Processing Vol. 27; no. 6; pp. 207 - 211
Main Authors Kawaguchi, Shoya, Kitamura, Daichi
Format Journal Article
LanguageEnglish
Japanese
Published Tokyo Research Institute of Signal Processing, Japan 01.11.2023
Japan Science and Technology Agency
Subjects
Algorithms
Amplitudes
Artificial neural networks
deep learning
mel-frequency cepstrum coefficient
Musical instruments
Spectrograms
timbre conversion
Online AccessGet full text

Cover

More Information
Summary:Timbre conversion of musical instrument sounds, utilizing deep neural networks (DNNs), has been extensively researched and continues to generate significant interest in the development of more advanced techniques. We propose a novel algorithm for timbre conversion that utilizes a variational autoencoder. However, this system must be capable of predicting the amplitude spectrogram from the melfrequency cepstrum coefficient (MFCC). This research aims to build a DNN-based decoder that utilizes the MFCC and time-frame-wise total amplitude as inputs to predict the amplitude spectrogram. Experiments conducted using a musical instrument sound dataset show that a decoder incorporating bidirectional long short-term memory yields accurate predictions of amplitude spectrograms.
ISSN:1342-6230
1880-1013
DOI:10.2299/jsp.27.207