Abstract
Developing decoders for imagined speech is crucial for individuals unable to produce audible speech, yet this process is challenging due to the lack of verifiable reference. This limitation highlights the need for innovative approaches to imagined speech decoding that can leverage the available verifiable reference from other types of speech production. To address this challenge, it is essential to understand the relationships between overt, mimed, and imagined speech. In our study, we utilized electrocorticography (ECoG) to examine neural correlates and decoder performance across these speech types during speech preparation. We observed high gamma activity (HGA) diverged earlier in the left hemisphere than in the right, suggesting an early stage of information processing for speech production. Also, HGA in overt and mimed speech showed similarity in the precentral and postcentral gyri, peaking at the end of speech preparation. In contrast, mimed and imagined speech demonstrated similarities in the superior temporal gyrus earlier in the process, also in HGA. These findings indicate that mimed speech shares neural signals with both overt and imagined speech in distinct spatial and temporal aspects, facilitating the development of imagined speech decoders for speech-impaired individuals.