Emotion Recognition on Speech Attributes Using Machine Learning

Modern day technology is controlled by A.I. technology and new A.I. Voice generators are popular in today's world. However, the words we speak aren't always the information we want to communicate and hence, the emotion of the speaker plays an important role in determining the exact information the speaker is trying to communicate. It is important to solve such a problem since many legal as well as professional recordings could be analyzed properly and accurate outcomes could be inferred. The dataset will undergo a preprocessing which will give us a clean, no noise dataset which further will get split into two different parts, test and train. SVM, MLP, RF, DT algorithms were applied on the dataset to get insights on the information given to us. Speaker vocal tract information, represented by Mel-frequency cepstral coefficients (MFCC), was extracted from the audio samples in the training dataset. Pitch, Short Term Energy (STE), and MFCC coefficients of audio samples in anger, happiness, fear, sadness like emotions were obtained. Then the nearest neighbor of unknown pitch, STE, and MFCC comparison for unknown data will decide the actual information on the data. The outcome must be validated and inferred for further references before deployment classifier algorithm came out to be most suitable for speech emotion recognition.