Award Date

1-1-1992

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

Number of Pages

75

Abstract

The Linear Prediction Analysis is one of the popular methods of processing speech. But it has problems in estimating the vocal tract characteristics of voiced sounds uttered by females and children. This is because the conventional linear prediction method assumes that all the sample values in each analysis frame are to be approximated by a linear combination of a definite number of the previous samples whether the previous samples include excitation periods or not. Also, the Linear Prediction analysis is easily affected by source excitation; The vocal tract characteristics of signals of short pitch period can be estimated more accurately by the Sample Selective Linear Prediction (SSLP). The first stage of a SSLP analysis is the conventional linear predictive analysis and in the second stage, only those samples which are under a specified threshold are used for further analysis; This work outlines a numerically stable algorithm for performing the SSLP using the Autocorrelation method. (Abstract shortened by UMI.).

Keywords

Analysis; Linear; Predictive; Sample; Selective; Signal; Speech

Controlled Subject

Electromagnetism; Artificial intelligence

File Format

pdf

File Size

1505.28 KB

Degree Grantor

University of Nevada, Las Vegas

Language

English

Permissions

If you are the rightful copyright holder of this dissertation or thesis and wish to have the full text removed from Digital Scholarship@UNLV, please submit a request to digitalscholarship@unlv.edu and include clear identification of the work, preferably with URL.

Identifier

https://doi.org/10.25669/6cfw-cbr8


Share

COinS