Speech Emotion Recognition for a Tribal Language using Machine Learning Methods

Emotion is the most powerful tool for communicating ideas and behaviours to another person. Speech-emotion recognition is essential for comprehending another person’s inner workings since human communication mostly depends on the capacity to read one another’s emotions. The suggested intelligent system may perform better by classifying emotions by integrating machine learning classifiers with time and frequency-based acoustic features. The KUI is a tribal dataset with limited resources used in this study. Speech may be used to extract emotion through its temporal and spectral properties. Multi-layer perceptron (MLP), Random Forest (RF), Support Vector Machine (SVM), and Decision Tree (DT) classifiers are used to classify emotions from the supplied wave signal. These models have been trained to recognize theseସଡାଙଗ (Angry), ରେହା (Happy), ଆଜ (Fear), ବକାଲ (Sad), ବଜାର (Disgust), ଡେକ (Surprise)). The authors measured Accuracy, Precision, Recall, and F1-Score to determine which model is optimal for the research. Among the four models, the SVM model has the highest accuracy of 94% and has the best precision rate, recall rate, and F1-Score forଆଜ(Fear) emotion.