A State Space Model for Decoding Auditory Attentional Modulation from MEG in a Competing Speaker Environment