Recursive short-time signal analysis /
Short-time signal analysis is a special tool for adaptive estimation, which estimates time-varying features of non-stationary signals or systems by using a slidingwindow to localize the data, then applying stationary estimation to the localized data to generate a local estimate. Short-time signal a...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1995.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=739663411&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | Short-time signal analysis is a special tool for adaptive estimation, which estimates time-varying features of non-stationary signals or systems by using a slidingwindow to localize the data, then applying stationary estimation to the localized data to generate a local estimate. Short-time signal analysis is a fundamental problem involved in adaptive signal processing, such as adaptive filtering, adaptive system identification, time-frequency/time-scale(wavelet) analysis, filter bank design, etc. Recursive short-time signal analysis is an efficient method to implement the short-time signal analysis, which estimates a local feature at a new time by updating those previously obtained ones with a new coming data. Windows suitable for this recursive implementation, called recursive windows, are very limited. Efficient recursive windows with desirable properties are highly demanded, especially for real-time processing. A rigorous definition of short-time signal analysis is given in Chapter II. Chapters III through VIII constitute the kernel part of this dissertation, in which a set of recursive windows and corresponding recursive procedures for short-time signal analysis are developed. These recursive windows include real-pole windows, complexpole windows, a cosine-wave window, and a half-sine-wave window. The structures, properties, and design procedures of these windows are carefully investigated. The recursive procedure of short-time signal analysis corresponding to each of these windows is derived in detail. The efficiencies of these procedures are also discussed. It is shown that complex-pole windows require exactly the same computation as the real-pole windows of the same order while their shapes are more desirable, implying better estimation is obtained at no cost. The recursive cosine-wave window and half-sine-wave window are bell-shaped: absolutely symmetric, single-peaked, and smoothly tapered off, therefore are among the ideal windows. The cosine-wave window and the half-sine-wave window require computation similar to the third order and the second order all-pole windows respectively for Short-Time Fourier Trans form (STFT). For average-based adaptive estimation, both of the two symmetric windows require only one multiplication operation coupled with several addition operations to update a new estimate, which is more efficient than all-pole windows. As compared with Amin's recursive procedure for both a Hamming window and a Hanning window, the recursive procedures of the cosine-wave window and the halfsine-wave window require about 50% and 40% respectively of the computation by Amin's method for STFT analysis, and 1/9 for average-based adaptive estimation. Chapters IX and X constitute the application part of this dissertation. A set of recursive implementations for STFT analysis by using the new recursive windows developed in this dissertation are presented. The structures of these new STFT implementations can be characterized by recursive computation along time and parallel computation for different frequencies. Based on this structure, a new time-frequency distribution possessing zoom-in and zoom-out capability is proposed. This new-time frequency distribution is very similar to that of STFT with one exception that a frequency dependent window is used. |
|---|---|
| Item Description: | Vita. "Major Subject: Electrical Engineering". |
| Physical Description: | xviii, 150 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilms Inc. |
| Bibliography: | Includes bibliographical references: pages 139-142. |