Abstract
A method for unseen emotion class recognition comprises: receiving an utterance sample to be tested with an emotion recognition model; calculating an embedding of the sample to be tested of the utterance sample to be tested with an encoder; calculating a first distance metric between the embedding of the sample to be tested and a representation of a first registered emotion category, and calculating a second distance metric between the embedding of the sample to be tested and a representation of a second registered emotion category, wherein the second registered emotion category is not included in a plurality of basic emotion categories; determining an emotion category of the utterance sample to be tested according to the first distance metric and the second distance metric.