Selected Features for Classifying Environmental Audio Data with Random Forest

ABSTRACT

Environmental audio classification has been the focus of research in the area of speech recognition. It is difficult to find an optimal classifier and select the optimal features from various features extracted from environmental audio data. Random forest is a powerful machine learning classifier compared to other conventional pattern recognition techniques. In this paper, the performance of the Random Forest classifier and the selected features model for environmental audio classification is explored. The comparison and analysis of classification results, obtained by employing the Bagging, AdaBoost, and Random Forest for environmental audio data, are given. The experiments carry out the assessment and selection of importance of variables by means of GINI Index. The results show that the Random Forest method can effectively improve the performance of environmental audio data classification even under the fewer training examples. The classification model, built from the selected features, obtains better performance in both accuracy and efficiency than that of all features for environmental audio data.

Keywords: