Abstract: The invention relates to a computer implemented method for real-time emotion recognition from a real-time audio signal. The method includes transcribing, into text, an audio speech signal contained in the audio signal by an automatic speech recognition model, and computing, by a speech representation model, a joint representation vector corresponding to a joint representation of the speech as a function of the speech signal and the text. The method also include computing, by an emotion prediction model, an emotion embedding vector as a function of the joint representation vector, and mapping the emotion in at least one emotional frame, according to the emotion embedding vector, by an emotion mapping model. The invention further relates to a computer program and a device implementing such a method.