In both datasets speakers do not repeat between test and train sets!
Path (on asya-5): /media/storage_2/data/processed/emo_audio/emo_audio_PP_threshold_2_and_other_from_relabelled_threshold_3
Class counts:
Train dataset: happiness: 1583 anger: 776 sadness: 1084 other: 6970
Test dataset: happiness: 141 anger: 141 sadness: 141 other: 141