Negative emotions in the target speaker’s voice enhance speech recognition under “cocktail-party” environments
