From pronounced to imagined: improving speech decoding with multi-condition EEG data Academic Article in Scopus uri icon