Signal Estimation from Short-Time Spectral Magnitude.
Abstract
This thesis shows that in many practical situations the processing of a discrete-time signal can be accomplished using only the magnitude of its short-time spectrum. Mild restrictions on the signal and on the analysis window of the short-time spectrum are shown to be sufficient for unique signal representation with the short-time spectral magnitude. Furthermore, various algorithms are developed which reconstruct the signal from appropriate samples of the short-time spectral magnitude. Some of these algorithms are designed to obtain signal estimates from the processed short-time spectral magnitude, which generally does not have a valid short-time structure. These algorithms are successfully applied to the time-scale modification and noise reduction problems in speech processing. However, the results presented here have similar potential for other application areas, including those with multidimensional signals. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1982
- Accession Number
- ADA119528
Entities
People
- Syed Hamid Nawab
Organizations
- Massachusetts Institute of Technology