Abstract:
This master thesis gives a way to analyze speech wavefore in order to store the speech at a lower bit rate ratner than the classical ways. During the analysis Linear Predictive Coding is used where speech is represented by some parameters such as energy, pitch and reflection coefficients. Energy parameter is based on residual energy. In order to determine pitch period Parallel Processing method is used. ihe reflection coefficients are calculated by using Gueguen Le-Ruex algorithm. These algorithms are applied to the isolated TurKish words. These calculated parameters are then fed to a speech synthesizer where isolated words are connected in order to form sentences. The results and observations are illustrated.
