A computational framework for speech emotion recognition in case of multisource data