# Singing Voice Synthesis Using Deep Autoregressive Neural Networks for Acoustic Modeling

arXiv: https://arxiv.org/abs/1906.08977 · HTML: https://mlsanity.com/paper/1906.08977/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 21 Jun 2019.

## Built on by (most cited first, 7 of 7)

- [Sinsy: A Deep Neural Network-Based Singing Voice Synthesis System](https://mlsanity.com/paper/2108.02776/) (2021)
- [SingAug: Data Augmentation for Singing Voice Synthesis with Cycle-consistent Training Strategy](https://mlsanity.com/paper/2203.17001/) (2022)
- [NNSVS: A Neural Network-Based Singing Voice Synthesis Toolkit](https://mlsanity.com/paper/2210.15987/) (2022)
- [DeepA: A Deep Neural Analyzer For Speech And Singing Vocoding](https://mlsanity.com/paper/2110.06434/) (2021)
- [Hierarchical Generative Modeling of Melodic Vocal Contours in Hindustani Classical Music](https://mlsanity.com/paper/2408.12658/) (2024)
- [Pitch Preservation In Singing Voice Synthesis](https://mlsanity.com/paper/2110.05033/) (2021)
- [Singing Voice Synthesis Based on a Musical Note Position-Aware Attention Mechanism](https://mlsanity.com/paper/2212.13703/) (2022)

## Abstract (arXiv)

This paper presents a method of using autoregressive neural networks for the acoustic modeling of singing voice synthesis (SVS). Singing voice differs from speech and it contains more local dynamic movements of acoustic features, e.g., vibratos. Therefore, our method adopts deep autoregressive (DAR) models to predict the F0 and spectral features of singing voice in order to better describe the dependencies among the acoustic features of consecutive frames. For F0 modeling, discretized F0 values are used and the influences of the history length in DAR are analyzed by experiments. An F0 post-processing strategy is also designed to alleviate the inconsistency between the predicted F0 contours and the F0 values determined by music notes. Furthermore, we extend the DAR model to deal with continuous spectral features, and a prenet module with self-attention layers is introduced to process historical frames. Experiments on a Chinese singing voice corpus demonstrate that our method using DARs can produce F0 contours with vibratos effectively, and can achieve better objective and subjective performance than the conventional method using recurrent neural networks (RNNs).