Výsledky vyhledávání - "FUJITA, Kenichi"

Report

Lightweight Zero-shot Text-to-Speech with Mixture of Adapters

Autor: Fujita, Kenichi, Ashihara, Takanori, Delcroix, Marc, Ijima, Yusuke

The advancements in zero-shot text-to-speech (TTS) methods, based on large-scale models, have demonstrated high fidelity in reproducing speaker characteristics. However, these models are too large for practical daily use. We propose a lightweight zer

Externí odkaz: http://arxiv.org/abs/2407.01291

Zobrazit plný text záznamu

Report

Speech Rhythm-Based Speaker Embeddings Extraction from Phonemes and Phoneme Duration for Multi-Speaker Speech Synthesis

Autor: Fujita, Kenichi, Ando, Atsushi, Ijima, Yusuke

Publikováno v: IEICE TRANSACTIONS on Information and Systems 107.1 (2024): 93-104

This paper proposes a speech rhythm-based method for speaker embeddings to model phoneme duration using a few utterances by the target speaker. Speech rhythm is one of the essential factors among speaker characteristics, along with acoustic features

Externí odkaz: http://arxiv.org/abs/2402.07085

Zobrazit plný text záznamu

Report

Noise-robust zero-shot text-to-speech synthesis conditioned on self-supervised speech-representation model with adapters

Autor: Fujita, Kenichi, Sato, Hiroshi, Ashihara, Takanori, Kanagawa, Hiroki, Delcroix, Marc, Moriya, Takafumi, Ijima, Yusuke

The zero-shot text-to-speech (TTS) method, based on speaker embeddings extracted from reference speech using self-supervised learning (SSL) speech representations, can reproduce speaker characteristics very accurately. However, this approach suffers

Externí odkaz: http://arxiv.org/abs/2401.05111

Zobrazit plný text záznamu

Report

Zero-shot text-to-speech synthesis conditioned using self-supervised speech representation model

Autor: Fujita, Kenichi, Ashihara, Takanori, Kanagawa, Hiroki, Moriya, Takafumi, Ijima, Yusuke

Publikováno v: 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW), 2023, pp. 1-5

This paper proposes a zero-shot text-to-speech (TTS) conditioned by a self-supervised speech-representation model acquired through self-supervised learning (SSL). Conventional methods with embedding vectors from x-vector or global style tokens still

Externí odkaz: http://arxiv.org/abs/2304.11976

Zobrazit plný text záznamu