Scholarlink

<p>Computation of the S-transform (ST) of a time series results in a 2-D complex time-frequency (t-f) representation that provides information about the local Fourier spectrum at every instant of time. The magnitude and phase spectra of the ST output provide quantitative and visual information about the local spectral content.The large computational and storage requirements of the ST place limitations on the usage of the ST. A recently developed compact S-transform (cST) points to an alternate method of computing the ST with significant economies in computation time and storage. With the cST, only a select fraction of the voices of ST are computed. Further significant economies are achieved through intra-voice decimation of the low frequency voices. The full ST matrix is then computed with interpolation. The resulting hybrid ST (hST) is in good agreement with the conventional direct computation of the ST. The improvements in computation speed and storage of the hST are more apparent with increase in the length of the time series.</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

Fast Hybrid Computation of the S-transform (ST) Through the Compact ST

  • 期刊论文

  • 发表日期：

    2026年4月30日

查看全文
  • Subrat Kumar Acharya,
  • Pyari Mohan Pradhan,
  • Lalu Mansinha

摘要

Computation of the S-transform (ST) of a time series results in a 2-D complex time-frequency (t-f) representation that provides information about the local Fourier spectrum at every instant of time. The magnitude and phase spectra of the ST output provide quantitative and visual information about the local spectral content.The large computational and storage requirements of the ST place limitations on the usage of the ST. A recently developed compact S-transform (cST) points to an alternate method of computing the ST with significant economies in computation time and storage. With the cST, only a select fraction of the voices of ST are computed. Further significant economies are achieved through intra-voice decimation of the low frequency voices. The full ST matrix is then computed with interpolation. The resulting hybrid ST (hST) is in good agreement with the conventional direct computation of the ST. The improvements in computation speed and storage of the hST are more apparent with increase in the length of the time series.