Acoustic feature selection and classification of emotions in speech using a 3D continuous emotion model