使用向量量化技術來達成語音轉換

Using Vector Quantization To Achieve Voice Conversion

吳達懿(Da-Yi Wu)

指導教授 ： 呂育道

共同指導教授 ： 李宏毅(Hung-yi Lee)

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202002653

摘要

本論文提出了一套基於向量量化的深度語者轉換模型。 與此同時，本文對此方法與其他現存的方法做了多種客觀、主觀的評估。結果顯示，本文所提出的方法在流利度以及語者相似度上都比現存的方法優秀。

關鍵字

語音轉換 ； 向量量化 ； 深度學習

並列摘要

This thesis proposes a vector quantization-based voice conversion approach. The objective and the subjective evaluations show that the proposed method performs better than other existing approaches in both audio naturalness and speaker similarity

並列關鍵字

voice conversion ； vector quantization ； deep learning

參考文獻

[1] J. Song, P. Kalluri, A. Grover, S. Zhao, and S. Ermon, “Learning controllable fair representations,” in Proceedings of 22nd International Conference on Artificial Intelligence and Statistics, April 16–18, 2019, Naha, Japan, pp. 2164–2173.

[2] F. Villavicencio and J. Bonada, “Applying voice conversion to concatenative singing-voice synthesis,” in 11th Annual Conference of the International Speech Communication Association, September 26–30, 2010, Makuhari, Japan, pp. 2162– 2165.

[3] E. Nachmani and L. Wolf, “Unsupervised singing voice conversion,” in 20th Annual Conference of the International Speech Communication Association, September 15– 19, 2019, Graz, Austria, pp. 2583–2587.

[4] S. H. Mohammadi and A. Kain, “Voice conversion using deep neural networks with speaker-independent pre-training,” in Spoken Language Technology Workshop, December 7–10, 2014, Lake Tahoe, pp. 19–23.

[5] M. Sahidullah and G. Saha, “Design, analysis and experimental evaluation of block based transformation in MFCC computation for speaker recognition,” Speech Commun., vol. 54, no. 4, pp. 543–565, 2012.