Formant set as a main parameter for recognizing vowels of the Uzbek language

The article highlights the applications of the algorithm for speech signal recognition using a formant set. A description of the formation of the speech signal by the human articulating apparatus is given. The influence of the human speech apparatus organs on the formant parameter is also described. The Linear Prediction Method (LPC) is used for determining the formant set in order to identify the peaks on the spectral diagram of the original signal. The paper presents the formant set for the Uzbek vowels pronounced by several speakers of the same age category with different genders after cochlear surgery. Based on the formant set graphs are constructed and classification problems over them are solved using machine learning algorithms such as K-Nearest Neighbors, Random Forest, Support Vector Machines. The scientific paper consists of a comparative analysis of the three main classification methods on such a priority parameter as recognition accuracy in order to simplify the process of speech signal recognition. The results of the classification revealed that the recognition of Uzbek vowel letters on the basis of the formant set is achieved using KNN and is 95%.

Hali tarjima qilinmagan