Scholarlink

<p>This research delineates the development of a hybrid architecture for text-to-speech (TTS) synthesis, termed the Efficient Speech Synthesizer (ESS). ESS leverages an end-to-end training paradigm to holistically optimize all the parameters, thereby facilitating the synthesis of high-fidelity speech with remarkable efficiency, irrespective of sentence length. The architecture integrates state-of-the-art feed-forward methodologies, including variational autoencoders and linear predictive coding techniques, which impose monotonic constraints on sequence alignment while incurring negligible computational overhead. The primary objective of this study is to engineer a TTS model characterized by minimal complexity and latency, rendering it suitable for deployment across a broad spectrum of computational platforms, ranging from high-performance systems to resource-constrained IoT devices and smartphones.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Efficient Speech Synthesizer: A Hybrid Monotonic Architecture for Text-to-speech via VAE & LPC-Net with Independent Sentence Length

  • 期刊论文

  • 发表日期：

    2025年3月25日

查看全文
  • Naveen Kumar Nallabala,
  • Balamurugan Souprayen,
  • Maruthamuthu Ramasamy,
  • Seshu Kumar Penumarti,
  • Nuthan Chandu Navuluri

摘要

This research delineates the development of a hybrid architecture for text-to-speech (TTS) synthesis, termed the Efficient Speech Synthesizer (ESS). ESS leverages an end-to-end training paradigm to holistically optimize all the parameters, thereby facilitating the synthesis of high-fidelity speech with remarkable efficiency, irrespective of sentence length. The architecture integrates state-of-the-art feed-forward methodologies, including variational autoencoders and linear predictive coding techniques, which impose monotonic constraints on sequence alignment while incurring negligible computational overhead. The primary objective of this study is to engineer a TTS model characterized by minimal complexity and latency, rendering it suitable for deployment across a broad spectrum of computational platforms, ranging from high-performance systems to resource-constrained IoT devices and smartphones.