Experimentální výzkum okamžité selekce pozornosti v prostředí s více mluvčími s využitím EEG.

Abstract

Recently it has been shown to be possible to ascertain the target of a subject´s attention in a cocktail party environment from single-trial (~60 s) electroencephalography (EEG) data. Specifically, this was shown in the context of a dichotic listening paradigm where subjects were cued to attend to a story in one ear while ignoring a different story in the other and were required to answer questions on both stories. This paradigm resulted in a high decodingaccuracy that correlated with task performance across subjects. Here, we extend this finding by showing that the ability to accurately decode attentional selection in a dichotic speech paradigm is robust to the particular attention task at hand. Subjects attended to one of two dichotically presented stories under four task conditions. These conditions required subjects to 1) answer questions on the content of the attended story, 2) detect vibrato targets in the voice of the attended speaker 3) answer questions and detect vibrato targets, and 4) detect target words in the attended stream. All four tasks led to high decoding-accuracy (~90%). In addition, we show that detection of the attended vibrato target evoked a distinctive event related potential (ERP). When incorporating these ERPs in our classifier, we were able to boost the average decoding-accuracy to 99%. These results offer new possibilities for developing user-friendly brain computer interfaces (BCIs). Keywords: EEG, Brain-computer interface, Cocktail party effect, Multi-speaker environment, Linear modelling, Auditory attention

Recently it has been shown to be possible to ascertain the target of a subject´s attention in a cocktail party environment from single-trial (~60 s) electroencephalography (EEG) data. Specifically, this was shown in the context of a dichotic listening paradigm where subjects were cued to attend to a story in one ear while ignoring a different story in the other and were required to answer questions on both stories. This paradigm resulted in a high decoding accuracy that correlated with task performance across subjects. Here, we extend this finding by testing if the ability to accurately decode attentional selection in a dichotic speech paradigm is robust to the particular attention task at hand. Subjects attended to one of two dichotically presented stories under four task conditions. These conditions required subjects to 1) answer questions on the content of the attended story, 2) detect irregular frequency fluctuations in the voice of the attended speaker 3) answer questions and detect frequency fluctuations, and 4) detect target words in the attended stream. Linear modelling techniques are used to generate Temporal Response Functions and analyse the neural response to low and high level tasks. Backward linear modelling is applied to reconstruct the stimulus and decode attentional selection for each subject. Decoding accuracy will be increased by classifying single event related potentials from both the attended and the unattended speech stream, using machine learning techniques. Decoding auditory attentional selection in real time, offers new possibilities for creating user-friendly brain computer interfaces (BCIs). In addition, it contributes to basic understanding of how the human brain solves the cocktail party problem and presents a new method of studying attention deficits in elderly people.