US20220130414A1 - SELECTION OF SPEECH SEGMENTS FOR TRAINING CLASSIFIERS FOR DETECTING EMOTIONAL VALENCE FROM INPUT SPEECH SIGNALS