Peking Opera Synthesis via Duration Informed Attention Network

Wu, Yusong; Li, Shengchen; Yu, Chengzhu; Lu, Hanqing; Weng, Chao; Zhang, Liqiang; Yu, Dong

doi:10.48550/arxiv.2008.03029

Cited by 4 publications

(3 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The model can generate high-fidelity and diverse songs with a coherence of several minutes. Yusong et al [233] even explored the composition of Peking Opera singing voice based on the Duration Informed Attention Network (DurIAN) framework.…”

Section: Singing Voice Synthesis(svs)mentioning

confidence: 99%

A Comprehensive Survey on Deep Music Generation: Multi-level Representations, Algorithms, Evaluations, and Future Directions

Ji,

Luo,

Yang

2020

Preprint

View full text Add to dashboard Cite

The utilization of deep learning techniques in generating various contents (such as image, text, etc.) has become a trend. Especially music, the topic of this paper, has attracted widespread attention of countless researchers.The whole process of producing music can be divided into three stages, corresponding to the three levels of music generation: score generation produces scores, performance generation adds performance characteristics to the scores, and audio generation converts scores with performance characteristics into audio by assigning timbre or generates music in audio format directly. Previous surveys have explored the network models employed in the field of automatic music generation. However, the development history, the model evolution, as well as the pros and cons of same music generation task have not been clearly illustrated. This paper attempts to provide an overview of various composition tasks under different music generation levels, covering most of the currently popular music generation tasks using deep learning. In addition, we summarize the datasets suitable for diverse tasks, discuss the music representations, the evaluation methods as well as the challenges under different levels, and finally point out several future directions.

show abstract

Section: Singing Voice Synthesis(svs)mentioning

confidence: 99%

A Comprehensive Survey on Deep Music Generation: Multi-level Representations, Algorithms, Evaluations, and Future Directions

Ji,

Luo,

Yang

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…As sequence-to-sequence (Seq2Seq) models have become the predominant architectures in neural-based TTS, state-of-the-art SVS systems have also adopted the encoder-decoder methods and showed improved performance over simple network structure (e.g., DNN, CNN, RNN) [17][18][19][20][21][22][23]. In these methods, the encoders and decoders vary from bi-directional Long-Short-Term Memory units (LSTM) to multi-head self-attention (MHSA) based blocks.…”

Section: Introductionmentioning

confidence: 99%

Sequence-To-Sequence Singing Voice Synthesis With Perceptual Entropy Loss

Shi

Guo

Huo

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

The neural network (NN) based singing voice synthesis (SVS) systems require sufficient data to train well and are are prone to over-fitting due to data scarcity. However, we often encounter data limitation problem in building SVS systems because of high data acquisition and annotation cost, . In this work, we propose a Perceptual Entropy (PE) loss derived from a psycho-acoustic hearing model to regularize the network. With a one-hour open-source singing voice database, we explore the impact of the PE loss on various mainstream sequence-to-sequence models, including the RNN-based, transformer-based, and conformer-based models. Our experiments show that the PE loss can mitigate the over-fitting problem and significantly improve the synthesized singing quality reflected in objective and subjective evaluations.

show abstract

“…The parameters include equalisation, compression, spatialization, accuracy, and precision. The existing used techniques include Convolutional Neural Network based Long Short Term Memory CNN-LSTM [18], Bidirectional based Long Short Term Memory (Bi-LSTM) [19], and Optical Music Recognition (OMR) [20].…”

mentioning

confidence: 99%

Application of LSTM Feature Extraction Model in the Construction of Opera Audio Database

Wei Zhou

2024

jes

View full text Add to dashboard Cite

Chinese traditional music plays a greater role in ideal society governed by rites and rituals. The musical culture of China is both ancient and modern. Throughout history, musicians have created several musical genres and used countless instruments. In order to appreciate Chinese traditional music, it is helpful to have a basic familiarity with three "keynotes": the classical traditional music of the elites; folk and ethnic music, and the most popular traditional Chinese instruments. In general, an opera audio database consists of audio recordings of opera performances. These recordings may be utilized for a variety of reasons, including research, education, and entertainment. To establish such a database, pertinent features must be extracted from the audio recordings to allow for effective storage, retrieval, and analysis of the data. Improvements in audio signal processing over the last several years have qualified for creating feature extraction methods that improve the quality and accuracy of the retrieved features. We first gather the audio recordings of opera performances using an appropriate recording device to build an opera audio database. Data preprocessing then eliminates noise and other undesired signals from the audio recordings to enhance the data quality. After that, to extract useful features from the audio recordings that may be utilized to define the sound's rates, the Long Short Term Memory (LSTM) technique is used. The components must then be organized and stored in a database for fast retrieval and analysis once extracted. A suitable database management system may be used to build the database. The database should be created to make it simple to search and filter the data using different criteria. The study results stated that building an opera audio database using novel feature extraction approaches is a challenging process that requires an understanding of opera music, database administration, and audio signal processing. The experimental results stated that the proposed model has provided an accuracy of 96%.

show abstract

Peking Opera Synthesis via Duration Informed Attention Network

Cited by 4 publications

References 24 publications

A Comprehensive Survey on Deep Music Generation: Multi-level Representations, Algorithms, Evaluations, and Future Directions

A Comprehensive Survey on Deep Music Generation: Multi-level Representations, Algorithms, Evaluations, and Future Directions

Sequence-To-Sequence Singing Voice Synthesis With Perceptual Entropy Loss

Application of LSTM Feature Extraction Model in the Construction of Opera Audio Database

Contact Info

Product

Resources

About