Μελέτη στατιστικών χαρακτηριστικών (pdf, power density spectrum, autocorrelation) σημάτων φωνής με χρήση του πακέτου λογισμικού Matlab
Study of statistical features (pdf, power density spectrum, autocorrelation) of voice signals using the software package Matlab
Μητράκα, Σοφία Χρύσα
SubjectΕπεξεργασία σημάτων ; Signal processing -- Digital techniques -- Data processing ; Signal processing -- Data processing ; MATLAB
Through this work an attempt was made to study some basic statistical characteristics of voice signals, which is necessary for subsequent processing. Such statistical characteristics is the probability density function (pdf), the power density spectrum and the autocorrelation function. In this thesis, first a description of the main features of the theory took place, on concepts such as random variables and the distribution function, and the probability density function. Also the main features of the spectrum power density and autocorrelation are presented, with reference to the basic definitions of concepts such series Fourier, deterministic and random / stochastic signals, Parseval's theorem and more. In addition, a presentation of the key characteristics of voice signals was made, focusing on time - frequency analysis of speech signals, the effect of the window length and key features in the analysis of such signals. In the experimental part, and using the Matlab software package, we studied three different statistical characteristics of voice signals, i.e. the probability density function, the power density spectrum and the autocorrelation function. The results of the implementation of code were presented for three different audio files with different attributes, and for two different values of the sampling frequency. The conclusions are particularly interesting, and the analysis can be done for a variety of audio signals with different characteristics, in order to acquire the researcher full knowledge of the basic characteristics of each signal before proceeding with further processing and analysis.