Özet:

In recent years, gender detection with many different fields of application has been a major problem in speech analysis. For gender detection, the various characteristics of sound data, such as curtains, media, frequency, are used. In this study, a character selection method is proposed based on metasecular optimization algorithms for gender detection from sound data. In the suggested method, the gender identification was made with artificial intelligence algorithms using selected and obtained features with the optimization algorithms of the set of features that will represent the sound data in the most appropriate way. Metaphysical optimization algorithms inspired by nature have been used with the ability to solve complex problems to make the choice of features from sound data. Particular Flow Optimization (PSO), Mormon Colony Optimization (KKO), Salp Flow Algorithm (SSA) and Baline Optimization Algorithm (BOA) are first modeled for the selection of features from sound data. A set of data open to general access has been used to measure the effectiveness of metaphor optimization algorithms. The performance for PSO, KKO, SSA and BOA’s selection of features is compared in terms of three different standards, including the conformity function value, the accuracy value and the number of selected features. The new data sets and the original data sets, obtained after the selection of the features with metasecular optimization algorithms, have been applied to Naive Bayes and Decision Tree algorithms. As a result of the analysis, this method, which uses metasecular optimization algorithms for the selection of characteristics, has seen an increase in the success rate in the results obtained by Naive Bayes and the Decision Tree algorithms.

Anahtar Kelimeler: