Detection of Emotion cues from Tamil Speech signals

Human Machine interface is an important requirement in many applications. Human Machine Interface (HMI) refers to the interaction between humans and machines, where machines are trained to understand and respond to human emotions in various situations using speech. In order to achieve this, speech recognition technology is utilized, which has different features that allow for the classification of emotions. Before the speech signals can be analyzed for emotion classification, they need to undergo pre-processing steps. This includes removing any irrelevant utterances, unvoiced signals, and noises. One effective technique for preprocessing is End Point Detection, which identifies the start and end points of each speech segment. Comparing the energy thresholds of different segments also helps in determining the boundaries of each speech segment. Once the speech signals have been pre-processed, features need to be extracted from them and compared with the trained speech coefficients. Several techniques can be used to extract features from real-time signals. One commonly used technique is Mel Frequency Cepstral Coefficients (MFCC), which represents the spectral envelope of the speech signal. Other features such as Spectral Centroid, Spectral Flatness, and Energy can also be extracted to better understand the characteristics of the speech signal. The features are classified using SVM, KNN, Neural Network and CNN to define emotions namely Happy, Sad, Angry and Neutral.