• Phonetics and Speech Sciences
• /
• v.16 no.1
• /
• pp.67-76
• /
• 2024

Recent advances in text-to-speech (TTS) technology have significantly improved the quality of synthesized speech, reaching a level where it can closely imitate natural human speech. Especially, TTS models offering various voice characteristics and personalized speech, are widely utilized in fields such as artificial intelligence (AI) tutors, advertising, and video dubbing. Accordingly, in this paper, we propose a one-shot multi-speaker TTS system that can ensure acoustic diversity and synthesize personalized voice by generating speech using unseen target speakers' utterances. The proposed model integrates a speaker encoder into a TTS model consisting of the FastSpeech2 acoustic model and the HiFi-GAN vocoder. The speaker encoder, based on the pre-trained RawNet3, extracts speaker-specific voice features. Furthermore, the proposed approach not only includes an English one-shot multi-speaker TTS but also introduces a Korean one-shot multi-speaker TTS. We evaluate naturalness and speaker similarity of the generated speech using objective and subjective metrics. In the subjective evaluation, the proposed Korean one-shot multi-speaker TTS obtained naturalness mean opinion score (NMOS) of 3.36 and similarity MOS (SMOS) of 3.16. The objective evaluation of the proposed English and Korean one-shot multi-speaker TTS showed a prediction MOS (P-MOS) of 2.54 and 3.74, respectively. These results indicate that the performance of our proposed model is improved over the baseline models in terms of both naturalness and speaker similarity.

• The Journal of the Convergence on Culture Technology
• /
• v.8 no.3
• /
• pp.469-475
• /
• 2022

To train the model of the deep learning-based single-speaker TTS system, a speech DB of tens of hours and a lot of training time are required. This is an inefficient method in terms of time and cost to train multi-speaker or personalized TTS models. The voice cloning method uses a speaker encoder model to make the TTS model of a new speaker. Through the trained speaker encoder model, a speaker embedding vector representing the timbre of the new speaker is created from the small speech data of the new speaker that is not used for training. In this paper, we propose a multi-speaker TTS system to which voice cloning is applied. The proposed TTS system consists of a speaker encoder, synthesizer and vocoder. The speaker encoder applies the d-vector technique used in the speaker recognition field. The timbre of the new speaker is expressed by adding the d-vector derived from the trained speaker encoder as an input to the synthesizer. It can be seen that the performance of the proposed TTS system is excellent from the experimental results derived by the MOS and timbre similarity listening tests.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.523-529
• /
• 2021

In this paper, we designed and developed an Emotional Speech Synthesis Markup Language (SSML) processor. Multi-speaker emotional speech synthesis technology that can express multiple voice colors and emotional expressions have been developed, and we designed Emotional SSML by extending SSML for multiple voice colors and emotional expressions. The Emotional SSML processor has a graphic user interface and consists of following four components. First, a multi-speaker emotional text editor that can easily mark specific voice colors and emotions on desired positions. Second, an Emotional SSML document generator that creates an Emotional SSML document automatically from the result of the multi-speaker emotional text editor. Third, an Emotional SSML parser that parses the Emotional SSML document. Last, a sequencer to control a multi-speaker and emotional Text-to-Speech (TTS) engine based on the result of the Emotional SSML parser. Based on SSML which is a programming language and platform independent open standard, the Emotional SSML processor can easily integrate with various speech synthesis engines and facilitates the development of multi-speaker emotional text-to-speech applications.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.33 no.9C
• /
• pp.691-696
• /
• 2008

A new vocoder of Code Excited Linear Predictive (CELP) based on Adaptive Multi Rate (AMR) 7.4kbit/s mode is proposed in this paper. The proposed vocoder achieves a better compression rate in an environment of Speaker Dependent Coding System (SDSC) and is efficiently used for systems, such as OGM(Outgoing message) and TTS(Text To Speech), which needs only one person's speech. In order to enhance the compression rate of a coder, a new Line Spectral Pairs(LSP) code-book is employed by using Centroid Neural Network (CNN) algorithm. In comparison with original(traditional) AMR 7.4 Kbit/s coder, the new coder shows 27% higher compression rate while preserving synthesized speech quality in terms of Mean Opinion Score(MOS).

• The Journal of the Acoustical Society of Korea
• /
• v.28 no.3
• /
• pp.290-297
• /
• 2009

This paper presents an improved adaptive multi-rate wideband (AMR-WB) algorithm for the efficient Text-To-Speech (TTS) database compression. The proposed algorithm includes unnecessary common bit-stream (CBS) removal and parameter delta coding combined with speaker-dependent huffman coding to reduce the required bit-rate without any quality degradation. We also propose lossy coding schemes to produce the maximum bit-rate reduction with negligible quality degradation. The proposed lossless algorithm including CBS removal can reduce bit-rate by 12.40% without quality degradation compared with the 12.65 kbps AMR-WB mode. The proposed lossy algorithm can reduce bit-rate by 20.00% with 0.12 PESQ degradation.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.2
• /
• pp.135-144
• /
• 2003

A method for the automatic segmentation of speech signals is described. The method is dedicated to the construction of a large database for a Text-To-Speech (TTS) synthesis system. The main issue of the work involves the refinement of an initial estimation of phone boundaries which are provided by an alignment, based on a Hidden Market Model(HMM). Multi-layer perceptron (MLP) was used as a phone boundary detector. To increase the performance of segmentation, a technique which individually trains an MLP according to phonetic transition is proposed. The optimum partitioning of the entire phonetic transition space is constructed from the standpoint of minimizing the overall deviation from hand labelling positions. With single speaker stimuli, the experimental results showed that more than 95% of all phone boundaries have a boundary deviation from the reference position smaller than 20 ms, and the refinement of the boundaries reduces the root mean square error by about 25%.