Scholarlink

<p>In our everyday routine, communication through speech is essential for interactions between individuals. Recently, TTS synthesis has attained special attention because of its promising applications in various aspects like education, accessibility, entertainment, and healthcare. The most popular TTS technique used in computer aided learning is Deep learning algorithms. The development of a Tamil TTS system using Deep Belief Network (DBN) with LASSO regularization is implemented in this paper. The LASSO regularization, improve the overall performance of Tamil TTS systems which offers high performance in feature extraction and robustness against over fitting. Hence the outcome of the proposed technique yields the best and accurate results compared to existing speech synthesis techniques. To improve the effectiveness of the system, a set of metrics, is calculated. It is shown that the proposed system outperforms existing Tamil TTS systems in terms of naturalness, intelligibility, and robustness. The simulation results achieve an improvement in accuracy of 98.6% compared to statistical based synthesis.</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 text to speech system for Tamil language based on deep learning approach

  • 期刊论文

  • 发表日期：

    2025年6月13日

查看全文
  • A. Femina Jalin,
  • J. Jaya Kumari

摘要

In our everyday routine, communication through speech is essential for interactions between individuals. Recently, TTS synthesis has attained special attention because of its promising applications in various aspects like education, accessibility, entertainment, and healthcare. The most popular TTS technique used in computer aided learning is Deep learning algorithms. The development of a Tamil TTS system using Deep Belief Network (DBN) with LASSO regularization is implemented in this paper. The LASSO regularization, improve the overall performance of Tamil TTS systems which offers high performance in feature extraction and robustness against over fitting. Hence the outcome of the proposed technique yields the best and accurate results compared to existing speech synthesis techniques. To improve the effectiveness of the system, a set of metrics, is calculated. It is shown that the proposed system outperforms existing Tamil TTS systems in terms of naturalness, intelligibility, and robustness. The simulation results achieve an improvement in accuracy of 98.6% compared to statistical based synthesis.