Year of Graduation
Machine Learning Methods Comparison: Spectrum and Sound
Fundamental and Computational Linguistics
The purpose of this paper is to compare the methods of machine learning for the classification of speech. The study was carried out for two types of data representation: audio data and spectrogram images. In the classification models, appropriate methods for feature extraction are used. The proposed method, based on the principles of image analysis, is not complex in implementation and shows a satisfactory result on the limited volume of Russian-language data of various properties.