• Phonetics and Speech Sciences
• /
• v.15 no.3
• /
• pp.69-74
• /
• 2023

This study introduces an innovative model for zero-shot voice conversion that utilizes the capabilities of HuBERT. Zero-shot voice conversion models can transform the speech of one speaker to mimic that of another, even when the model has not been exposed to the target speaker's voice during the training phase. Comprising five main components (HuBERT, feature encoder, flow, speaker encoder, and vocoder), the model offers remarkable performance across a range of scenarios. Notably, it excels in the challenging unseen-to-unseen voice-conversion tasks. The effectiveness of the model was assessed based on the mean opinion scores and similarity scores, reflecting high voice quality and similarity to the target speakers. This model demonstrates considerable promise for a range of real-world applications demanding high-quality voice conversion. This study sets a precedent in the exploration of HuBERT-based models for voice conversion, and presents new directions for future research in this domain. Despite its complexities, the robust performance of this model underscores the viability of HuBERT in advancing voice conversion technology, making it a significant contributor to the field.

• Phonetics and Speech Sciences
• /
• v.12 no.3
• /
• pp.55-63
• /
• 2020

Voice conversion can be applied to various voice processing applications. It can also play an important role in data augmentation for speech recognition. The conventional method uses the architecture of voice conversion with speech synthesis, with Mel filter bank as the main parameter. Mel filter bank is well-suited for quick computation of neural networks but cannot be converted into a high-quality waveform without the aid of a vocoder. Further, it is not effective in terms of obtaining data for speech recognition. In this paper, we focus on performing voice-to-voice conversion using only the raw spectrum. We propose a deep learning model based on the transformer network, which quickly learns the voice conversion properties using an attention mechanism between source and target spectral components. The experiments were performed on TIDIGITS data, a series of numbers spoken by an English speaker. The conversion voices were evaluated for naturalness and similarity using mean opinion score (MOS) obtained from 30 participants. Our final results yielded 3.52±0.22 for naturalness and 3.89±0.19 for similarity.

• Phonetics and Speech Sciences
• /
• v.6 no.4
• /
• pp.133-139
• /
• 2014

This paper describes several approaches to transform a speaker's individuality to another's individuality using frequency warping between bilingual formant frequencies on different language environments. The proposed methods are simple and intuitive voice conversion algorithms that do not use training data between different languages. The approaches find the warping function from source speaker's frequency to target speaker's frequency on formant space. The formant space comprises four representative monophthongs for each language. The warping functions can be represented by piecewise linear equations, inverse matrix. The used features are pure frequency components including magnitudes, phases, and line spectral frequencies (LSF). The experiments show that the LSF-based voice conversion methods give better performance than other methods.

• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.115-118
• /
• 2000

Voice conversion technology is essential for TTS systems because the construction of speech database takes much effort. In this paper. male-to-female voice conversion technology in Korean LPC TTS system has been studied. In general. the parameters for voice color conversion are categorized into acoustic and prosodic parameters. This paper adopts LSF(Line Spectral Frequency) for acoustic parameter, pitch period and duration for prosodic parameters. In this paper. Pitch period is shortened by the half, duration is shortened by $25\%, and LSFs are shifted linearly for the voice conversion. And the synthesized speech is post-filtered by a bandpass filter. The proposed algorithm is simpler than other algorithms. for example, VQ and Neural Net based methods. And we don't even need to estimate formant information. The MOS(Mean Opinion Socre) test for naturalness shows 2.25 and for female closeness, 3.2. In conclusion, by using the proposed algorithm. male-to-female voice conversion system can be simply implemented with relatively successful results.

• MALSORI
• /
• v.68
• /
• pp.65-74
• /
• 2008

A statistical parametric speech synthesis system based on the hidden Markov models (HMMs) has grown in popularity over the last few years, because it needs less memory and low computation complexity and is suitable for the embedded system in comparison with a corpus-based unit concatenation text-to-speech (TTS) system. It also has the advantage that voice characteristics of the synthetic speech can be modified easily by transforming HMM parameters appropriately. In this paper, we present experimental results of voice characteristics conversion using the HMM-based Korean speech synthesis system. The results have shown that conversion of voice characteristics could be achieved using a few sentences uttered by a target speaker. Synthetic speech generated from adapted models with only ten sentences was very close to that from the speaker dependent models trained using 646 sentences.

• MALSORI
• /
• no.45
• /
• pp.63-77
• /
• 2003

The purpose of voice color conversion is to change the speaker identity perceived from the speech signal. In this paper, we propose a new voice color conversion algorithm through the formant shifting and the spectrum-tilt modification in the frequency domain. The basic idea of this technique is to convert the positions of source formants into those of target speaker's formants through interpolation and decimation and to modify the spectrum-tilt by utilizing the information of both speakers' spectrum envelops. The LPC spectrum is adopted to evaluate the position of formant and the information of spectrum-tilt. Our algorithm enables us to convert the speaker identity rather successfully while maintaining good speech quality, since it modifies speech waveforms directly in the frequency domain.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.6
• /
• pp.71-76
• /
• 1999

In digital cellular communication systems, tandem coding occurs in communications between mobile phones with different speech codecs, resulting in poor voice quality, high computational load, and long transmission delay. In this paper, voice packet conversion technique is proposed to solve the tandem coding problems, and packet conversion algorithm from 13kbps QCELP to 8kbps QCELP is developed. Simulations using various speech data show that the proposed packet conversion method produces voice quality which is equivalent to that by the conventional tandem coding method with shorter transmission delay using about 33% computational load.

• The Journal of the Acoustical Society of Korea
• /
• v.41 no.3
• /
• pp.351-358
• /
• 2022

Recently, Generative Adversarial Networks (GAN) and Variational AutoEncoders (VAE) have been applied to voice conversion that can make use of non-parallel training data. Especially, Conditional Cycle-Consistent Generative Adversarial Networks (CC-GAN) and Cycle-Consistent Variational AutoEncoders (CycleVAE) show promising results in many-to-many voice conversion among multiple speakers. However, the number of speakers has been relatively small in the conventional voice conversion studies using the CC-GANs and the CycleVAEs. In this paper, we extend the number of speakers to 100, and analyze the performances of the many-to-many voice conversion methods experimentally. It has been found through the experiments that the CC-GAN shows 4.5 % less Mel-Cepstral Distortion (MCD) for a small number of speakers, whereas the CycleVAE shows 12.7 % less MCD in a limited training time for a large number of speakers.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2018.10a
• /
• pp.369-372
• /
• 2018

This paper suggests non-parallel-voice-conversion network conversing voice between unmapped voice pair as source voice and target voice. Conventional voice conversion researches used learning methods that minimize spectrogram's distance error. Not only these researches have some problem that is lost spectrogram resolution by methods averaging pixels. But also have used parallel data that is hard to collect. This research uses PPGs that is input voice's phonetic data and a GAN learning method to generate more clear voices. To evaluate the suggested method, we conduct MOS test with GMM based Model. We found that the performance is improved compared to the conventional methods.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.2
• /
• pp.713-725
• /
• 2022

Mostly, artificial intelligence does not show any definite change in emotions. For this reason, it is hard to demonstrate empathy in communication with humans. If frequency modification is applied to neutral emotions, or if a different emotional frequency is added to them, it is possible to develop artificial intelligence with emotions. This study proposes the emotion conversion using the Generative Adversarial Network (GAN) based voice frequency synthesis. The proposed method extracts a frequency from speech data of twenty-four actors and actresses. In other words, it extracts voice features of their different emotions, preserves linguistic features, and converts emotions only. After that, it generates a frequency in variational auto-encoding Wasserstein generative adversarial network (VAW-GAN) in order to make prosody and preserve linguistic information. That makes it possible to learn speech features in parallel. Finally, it corrects a frequency by employing Amplitude Scaling. With the use of the spectral conversion of logarithmic scale, it is converted into a frequency in consideration of human hearing features. Accordingly, the proposed technique provides the emotion conversion of speeches in order to express emotions in line with artificially generated voices or speeches.