The paper describes an experimental study on vocal emotion expression and recognition. Utterances expressing five emotions - happiness, anger, sadness, fear, and normal (unemotional) state - portrayed by thirty non-professional actors were recorded and then replayed to them for evaluation. The results on decoding emotions are consistent with earlier findings. The results on ability of humans to portray emotions and to recognize their own emotions in speech are presented. Computer algorithms for recognizing emotions in speech using ensembles of neural networks were developed and tested. Their accuracy is presented.