Ensemble of machine learning and acoustic segment model techniques for speech emotion and autism spectrum disorders recognition.