Scholarlink

<p>The paper studies a new modeling methodology in the short-time Fourier transform domain for speech signals. A robust speech enhancement technique is achieved by using an improved minima-controlled recursive averaging noise estimating technique and an optimally modified log-spectral amplitude estimator, where the spectral variance is estimated based on a stochastic volatility model. The suggested method based on the stochastic volatility model takes into account the heavy-tailed distribution of the short-time Fourier transform expansion coefficients and provides a plausible model to estimate their variances. The speech variances modeled are then estimated using the Bayesian method via the Markov chain Monte Carlo techniques. A comparison to the existing generalized autoregressive conditionally heteroscedastic modeling method is also carried out using different objective quality measures.</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

Stochastic volatility models for speech signals

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Fathima Jafna,
  • S. D. Krishnarani

摘要

The paper studies a new modeling methodology in the short-time Fourier transform domain for speech signals. A robust speech enhancement technique is achieved by using an improved minima-controlled recursive averaging noise estimating technique and an optimally modified log-spectral amplitude estimator, where the spectral variance is estimated based on a stochastic volatility model. The suggested method based on the stochastic volatility model takes into account the heavy-tailed distribution of the short-time Fourier transform expansion coefficients and provides a plausible model to estimate their variances. The speech variances modeled are then estimated using the Bayesian method via the Markov chain Monte Carlo techniques. A comparison to the existing generalized autoregressive conditionally heteroscedastic modeling method is also carried out using different objective quality measures.