An implementation of anger detection in speech signals

Abstract

In this paper, an emotion classification system based on speech signals is presented. The classifier can identify the most common emotions, namely anger, neutral, happiness and fear. The algorithm computes a number of acoustic features which are fed into the classifier based on a pattern recognition approach. The classification system is of potential benefit for ambient intelligence in which the emotional and physical states of a person should be known to the intelligence of the environment. Using such information, the environment can better support humans in their daily activities in accordance with their preferences.

Techniques in Pattern Recognition for School Bullying Prevention: Review and Outlook

“…The most common voice features include pitch [4][5][6][7][8][9], energy [4,5], intensity [6], formants [7], and MFCCs (Mel Frequency Cepstral Coefficients) [4,8,10,11].…”

“…There is a large number of classifiers for speech emotion recognition, e.g. K-Nearest Neighbor (K-NN) [4], Maximum Likelihood Bayes (MLB) [5], Artificial Neural Network (ANN) [6], MultiLayer Perceptron (MLP) [7], Hidden Markov Model (HMM) [11], Gaussian Mixture Model (GMM) [17], Support Vector Machine (SVM), Neural Network (NN) [18], Probabilistic Neural Network (PNN) [19], etc.…”

“…The databases used for training or recognizing include DES (Danish Emotional Speech) [5], SUSAS (Speech Under Simulated and Actual Stress) [5], EMO-DB (Berlin EMOtional DataBase) [20], ORI (Oregon Research Institute) [21], etc., which are in different languages. EMO-DB is the database most commonly used by researchers [4,12,22,23]. Some researchers also use real-life recordings as the databases [24,25].…”