Abstract
In this study, a new modeliing approach for reconstruction of speech signal by the Signature base Sequences is presented. The speech signal is defined as a linear combination of Signature Base Sequances. Signature Base Sequences generated from one, two and three-sound syllables of the Turkish language. Similar syllables have been omitted by tolerance of 10% for effective coding and then the number of syylables has been reduced by 30%. An attempt has been made to obtain speech signals by using the sets of syllables thus obtained. lt was found that the reduced number of syylable sets gave an error in the order of 0.0001 based on comparison of the speech signals. It was shown that this level of the error did not impair the comprehensibility of the speech signal. The aim of this study is to assign a reduced optimal minimum set of Signature Base Frequencies to Turkish sounds.