P300-数据集免费

资源介绍

Context [P300](https://www.ncbi.nlm.nih.gov/pubmed/1464675) dataset for eight healthy subjects. This dataset was produced using the standard 6x6 Donchin and Farewell P300 Speller Matrix, with an ISI of 0.125 ms. There are 7 words with 5 letters each. There are 10 intensification sequences per letter. The original procedure used 3 words for training, and tried to decode the remaining 4 words for testing. Subject Age Gender Right-handedness 1 56 M R 2 23 M L 3 21 F R 4 29 M R 5 20 F R 6 24 M R 7 29 M R 8 30 M L Content The most important structure is **data**. - data.X : EEG Matrix (8 channels) - data.y : Labels (1/2) - data.y_stim: Stimulation number: 1-6 rows, 7-12 cols - data.trial: Sample point where each of the 35 trials starts. - data.flash: Sample point where each flashing starts (sample point id, duration, stimulation, hit/nohit) Device: g.Tec g.Nautilus g.LadyBird, 250 Hz, notch filter to 50Hz, bandpass 0.1-30 Hz Acknowledgements We would like to thanks all the subjects that voluntarily accepted to participate in this experiment. We hope they have enjoyed the "Alfajor" that we gave them. Inspiration Try to decode the spelled word directly from the EEG matrix. There are 7 words of 5 letters each. Each letter is composed of 120 stimulations of the P300 matrix, 6 rows and 6 columns, ten times each. The goal is to decode the spelled words from the last 20 letters (4 words). As the ISI is very low, it is challenging to acquire good performances. Produced by the CiC, ITBA University, Buenos Aires, Argentina