Human Speech Emotion Classification Based on Bengali Language

Abstract

The research focuses on the task of human speech emotion recognition in the Bengali language. Due to the variety of ways that emotions may be communicated, it can be difficult to identify emotion from speech. Four different machine learning algorithms were used to classify speech emotions: Random Forest, SVM, CatBoost, and XGBoost. The dataset used for this research was collected from native Bengali speakers and consisted of speech samples expressing different emotions: anger, happiness and neutral. The speech samples were pre-processed to extract various features using MFCCs and LPC coefficients. The Random Forest algorithm, according to experimental findings, has the best accuracy of 70.42%. Regression and classification problems may be accomplished using the robust and flexible machine learning method random forest. And for an effective method of feature selection and provides a relatively high accuracy. These results demonstrate that Random Forest is a suitable algorithm for emotion recognition in Bengali speech. This research shows that machine learning algorithms can be used to effectively recognize emotions in Bengali speech. The highest accuracy was achieved using Random Forest, which suggests that it is a suitable algorithm for this task. Further research can be done to improve the performance of these algorithms by using more sophisticated feature extraction techniques or incorporating other modalities such as facial expressions or physiological signals.