Výsledky vyhledávání

Report

FastFit: Towards Real-Time Iterative Neural Vocoder by Replacing U-Net Encoder With Multiple STFTs

Autor: Jang, Won, Lim, Dan, Park, Heayoung

This paper presents FastFit, a novel neural vocoder architecture that replaces the U-Net encoder with multiple short-time Fourier transforms (STFTs) to achieve faster generation rates without sacrificing sample quality. We replaced each encoder block

Externí odkaz: http://arxiv.org/abs/2305.10823

Zobrazit plný text záznamu

Report

JETS: Jointly Training FastSpeech2 and HiFi-GAN for End to End Text to Speech

Autor: Lim, Dan, Jung, Sunghee, Kim, Eesung

In neural text-to-speech (TTS), two-stage system or a cascade of separately learned models have shown synthesis quality close to human speech. For example, FastSpeech2 transforms an input text to a mel-spectrogram and then HiFi-GAN generates a raw wa

Externí odkaz: http://arxiv.org/abs/2203.16852

Zobrazit plný text záznamu

Report

UnivNet: A Neural Vocoder with Multi-Resolution Spectrogram Discriminators for High-Fidelity Waveform Generation

Autor: Jang, Won, Lim, Dan, Yoon, Jaesam, Kim, Bongwan, Kim, Juntae

Most neural vocoders employ band-limited mel-spectrograms to generate waveforms. If full-band spectral features are used as the input, the vocoder can be provided with as much acoustic information as possible. However, in some models employing full-b

Externí odkaz: http://arxiv.org/abs/2106.07889

Zobrazit plný text záznamu

Report

Universal MelGAN: A Robust Neural Vocoder for High-Fidelity Waveform Generation in Multiple Domains

Autor: Jang, Won, Lim, Dan, Yoon, Jaesam

We propose Universal MelGAN, a vocoder that synthesizes high-fidelity speech in multiple domains. To preserve sound quality when the MelGAN-based structure is trained with a dataset of hundreds of speakers, we added multi-resolution spectrogram discr

Externí odkaz: http://arxiv.org/abs/2011.09631

Zobrazit plný text záznamu

Report

JDI-T: Jointly trained Duration Informed Transformer for Text-To-Speech without Explicit Alignment

Autor: Lim, Dan, Jang, Won, O, Gyeonghwan, Park, Heayoung, Kim, Bongwan, Yoon, Jaesam

We propose Jointly trained Duration Informed Transformer (JDI-T), a feed-forward Transformer with a duration predictor jointly trained without explicit alignments in order to generate an acoustic feature sequence from an input text. In this work, ins

Externí odkaz: http://arxiv.org/abs/2005.07799

Zobrazit plný text záznamu

Report

Convolutional Attention-based Seq2Seq Neural Network for End-to-End ASR

Autor: Lim, Dan

This thesis introduces the sequence to sequence model with Luong's attention mechanism for end-to-end ASR. It also describes various neural network algorithms including Batch normalization, Dropout and Residual network which constitute the convolutio

Externí odkaz: http://arxiv.org/abs/1710.04515

Zobrazit plný text záznamu

Dissertation/ Thesis

Characterization of thermal dissipation within integrated gate bipolar transistor (IGBT) layered packaging structure

Autor: Lim, Dan J.

Integrated Gate Bipolar Transistors (IGBTs) generally have a high output power and generate significant amounts of heat, which needs to be removed from the chip to ensure continued operation. Since IGBT chips are commonly mounted on a layered assembl

Externí odkaz: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.494895

Zobrazit plný text záznamu