Automatic speech emotion detection using hybrid of gray wolf optimizer and naïve Bayes

Over the past decade, automatic speech emotion detection has been a great challenge in the human–computer interaction area. Generally, individuals express their feelings explicitly or implicitly through words, facial expressions, gestures, or writing. Different datasets such as speech, text, and visuals are used to explore emotions. Here, seven emotions such as neutrality, happiness, sadness, fear, surprise, disgust, and anger are detected using speech signals. To perform speech emotion recognition, several datasets are available. SAVEE and TESS datasets are used here. In most of the earlier works, separate databases were used to identify emotions. But here, SAVEE and TESS databases are merged to create a new database and identified their emotions. Our main objective is to use this robust dataset to characterize their emotions. For this purpose, we have proposed a new machine learning algorithm. Initially, Mel-frequency cepstral coefficients are utilized to extract the features from the voice signal datasets. Finally, a hybrid of gray wolf optimizer and naïve Bayes machine learning algorithm was proposed for classification. From the results, our proposed classification algorithm provides better performance compared to current machine learning.