Imagined Speech Datasets Applying Traditional and Gamified Acquisition Paradigms

Recent computational advances have benefited brain-computer interfaces, but human factors have been continuously overlooked. Paradigm design directly impacts users' emotional state, thus impacting brain signal quality. This dataset provides electroencephalographic (EEG) data of 15 participants while performing imagined speech. Each participant performed two paradigms: a traditional paradigm, based on conventional design; and an experimental (gamified) paradigm, based on a video game. All files were pre-processed using EEGLAB. Raw and pre-processed data are provided.

Additionally, a set of questionnaires was applied to allow the study of the effects of each paradigm on each participant. First, age and sex were registered. Participants then answered the Internal Representations Questionnaire, which provides individual propensities to different kinds of internal representations. The measured representations are visual imagery, internal verbalization, and orthographic imagery. This questionnaire also provides a self-rating on manipulating mental representations. Participants also answered a Self-Assessment Manikin test before and after each paradigm, to establish emotional state, and a User Experience Questionnaire after each paradigm, to evaluate each paradigm across attractiveness, perspicuity, efficiency, dependability, stimulation, and novelty.

The Experimental Paradigm includes vocalized speech after every imagined speech instance. Each vocalized word was registered on a spreadsheet. Instances that contain errors are not recommended for training classification models.

Possible applications of this dataset include, but are not limited to: classifying between up to four different imagined speech words, identifying the differences between imagined speech and vocalized speech, and researching the impact of the acquisition paradigm on brain signals quality.