Recently, advances in voice-based multimedia applications and the Internet of Things (IoT) have led to the creation of large amounts of big data, containing patterns, correlations, and trends that reflect diverse aspects of human behaviour. Speech-based emotion recognition plays a pivotal role in improving human-machine interactions, emphasizing the significance of analyzing audio data from human conversations to extract emotional insights. Developing a robust emotion detection framework is essential and involves various modules, including speech-to-text conversion, feature extraction, feature selection, and classification of these features to determine emotions. Applications of audio processing technology are vital in improving human-computer interaction. In this study, three classification algorithms were used to evaluate the performance of emotion recognition (happiness, sadness, and anger) by feeding audio signals into the classifiers and comparing the results. The recognition accuracy using the One-R algorithm was (74%) with a processing time of 0.02 s. While the recognition accuracy achieved by the other two classifiers was Support Vector Machines SVM (64) with a processing time of 0.4 s and Random Forests RF (58%) with a processing time of 0.06 s. Through comparison between the classifiers, it was found that the (One-R) classifier provided the best performance and classification accuracy and the shortest time to implement.