{
"fetched_at": "2026-10-24T01:32:53Z",
"source_url": "https://www.kaggle.com/competitions/eeg-biometric-competition/data",
"pages": [
{
"name": "data-description",
"content": "## Enrollment, Calibration and Testing Set\nIn this competition, the M3CV dataset is splited into the Enrollment Set, Calibration Set, and Testing Set:\n\n1. Enrollment Set: the EEG data of session 1 from 95 subjects \n - For the participants to train the biometric model.\n2. Calibration Set: the EEG data of session 2 from 20 subjects\n - For the participants to observe the EEG data, analyze the commonalities and variability between different subjects, different days, and different paradigms, and conduct offline comparison between different prediction models.\n - Not all the EEG epoches in session 2 in these 20 subjects are provided in the calibration set. Part of the other EEG epoches has been arranged in testing set with subject ID unlabeled.\n3. Testing Set: the EEG data of session 2 from 86 subjects\n - Public leaderboard : around 10% of epochs in testing set\n - Private leaderboard : around 90% epochs in testing set\n - 11 intruders have been contained in the testing set, which does not appear in the enrollment set.\n - The information for which epoch belongs to the public leaderboard or the private leaderboard is hidden.\n\n\n\n\n\n\n## Data format and signal preprocessing\nAll epoches were arranged in the mat file,named epochxxxxxx.mat, with a 2D-matrix epoch_data （$65\\times 1000$ ）in it. Rows 1-64 correspond to 64 channels of EEG signal, row 65 corresponds to the marker information. xxxxxx are the random ID for each epoch. BrainAmp (brain products GmbH, Germany) was used for EEG recording with the original sampling rate 1000Hz. Each epoch are segmented from the preprocessed EEG signal, the preprocessing step including\n - Bad channel repairing\n - Rereference to TP9/TP10\n - Bandpass filter: 0.01-200Hz\n - ICA filter for artifact removing\n - Downsampling to 250Hz\n\n## Files\nThree zip files are provided with all the epoches used in this competition, which are\n\n* **Enrollment.zip** - the epochs in enrollment set\n* **Calibration.zip** - the epochs in calibration set\n* **Testing.zip** -the epochs in testing set\n* **Enrollment_Info.csv** -the label information for enrollment Set\n* **Calibration_Info.csv** -the label information for calibration Set\n* **Testing_Info.csv** -the label information for testing Set\n* **Baseline.zip** -matlab version code of the baseline methods\n* **Result_example_0907_181703.csv** -an example sumbission *.csv file \n \n## Columns\n\n* `Epoch_ID` - The ID of the epoch, \n - ranged from epoch000001 to epoch120000\n* `Subject_ID` - The ID of the subject\n - ranged from sub001 to sub106, or None\n - if usage is 1 or 2, the Subject_ID is the real ID of the subject\n - if usage is 3, the Subject_ID is None\n - if usage is 4, the Subject_ID is the ID that the subject claim he or she is.\n* `Session` - The number of the session, from 1 to 2\n* `Task` - The label of the Task, from 1 to 13\n - 1-EC\n - 2-EO\n - 3-VEP\n - 4-AEP\n - 5-SEP\n - 6-SSVEP\n - 7-SSAEP\n - 8-SSSEP\n - 9-P300（Target and Nontarget）\n - 10-FT\n - 11-RH\n - 12-LH\n - 13-SSVEP-SA\n* `Usage` - The label for the usage of the Epoch, ranged from 1 to 4\n - 1-Enrollment\n - 2-Calibration\n - 3-Identification\n - 4-Verification\n\n\n## Electrode\n* 1-Fp1\n* 2-Fp2\n* 3-F3\n* 4-F4\n* 5-C3\n* 6-C4\n* 7-P3\n* 8-P4\n* 9-O1\n* 10-O2\n* 11-F7\n* 12-F8\n* 13-T7\n* 14-T8\n* 15-P7\n* 16-P8\n* 17-Fz\n* 18-Cz\n* 19-Pz\n* 20-FC1\n* 21-FC2\n* 22-CP1\n* 23-CP2\n* 24-FC5\n* 25-FC6\n* 26-CP5\n* 27-CP6\n* 28-FT9\n* 29-FT10\n* 30-TP9\n* 31-TP10\n* 32-F1\n* 33-F2\n* 34-C1\n* 35-C2\n* 36-P1\n* 37-P2\n* 38-AF3\n* 39-AF4\n* 40-FC3\n* 41-FC4\n* 42-CP3\n* 43-CP4\n* 44-PO3\n* 45-PO4\n* 46-F5\n* 47-F6\n* 48-C5\n* 49-C6\n* 50-P5\n* 51-P6\n* 52-AF7\n* 53-AF8\n* 54-FT7\n* 55-FT8\n* 56-TP7\n* 57-TP8\n* 58-PO7\n* 59-PO8\n* 60-Fpz\n* 61-CPz\n* 62-POz\n* 63-Oz\n* 64-FCz\n* 65-Marker",
"isPublished": false,
"hasIsPublished": false,
"postTitle": "",
"hasPostTitle": false,
"mimeType": "",
"hasMimeType": false
},
{
"name": "Description",
"content": "\n# 1. Challenge Description\n\nIn the current digital information society, the biometrics technique is an essential tool in many personal and enterprise security systems. Biometrics is an important personal authentication or identification technique based on biological measurements, including physical, physiological, or behavioral characteristics. Compared with existing voice recognition, fingerprint scanning, and facial recognition, EEG-based biometrics has unique advantages (e.g., it is more difficult to be stolen, forged, or damage, and it must be detected in vivo). Hence, EEG-based biometrics is capable of providing a more secure biometric method for identity recognition.\n\nHowever, at present, EEG-based biometric technique is still in the exploratory stage. Current EEG-based biometric studies are mostly focused on a small number of samples and lack evaluation of multi-session and multi-task EEG recordings, which made the development of EEG-based biometric systems challenging with their uniqueness, permanence, and robustness. At the same time, there is no open EEG-based biometric competition and a suitable database. The lack of a unified test benchmark and platform seriously hinders the development of this field. To this end, we collected the M3CV (A **M**ulti-subject, **M**ulti-session, and **M**ulti-task Database for Investigation of EEG **C**ommonality and **V**ariability) dataset and launched an EEG-based biometric competition.\n\nCompared with the existing open dataset, this competition list four challenges as follows:\n\n1. Larger population size\n2. Multiple sessions of EEG recording\n3. Multiple tasks of EEG recording\n4. The intruders for the identification and verification\n\n# 2. Task Definition\nWith the 106 subjects, 2 sessions, and 14 types of EEG recording in M3CV database, we segmented the EEG signal and sorted them out of order. The EEG-based biometric competition was launched by focusing on the problems of personal identification and verification. Based on the enrollment dataset, the participant needs to train the prediction model for the biometric system. After the model has been trained, the participant would be asked to complete the following two prediction tasks:\n- **Identification**: Determine whether a given EEG signal comes from a subject in the Enrollment Set, and further determine the subject ID if he or she is not an intruder.\n- **Verification**: Determine whether a given EEG signal comes from a certain subject.\n\n\n# 3. Dataset Provider\n**HUANG Gan** Assistant Professor\n**LIANG Zhen** Assistant Professor\n**ZHANG Zhiguo** Professor\n![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F10539915%2F347d5e94941e0daad9e85f2e406bbaa8%2F0101151s.png?generation=1662828199475233&alt=media\")
\n",
"isPublished": false,
"hasIsPublished": false,
"postTitle": "",
"hasPostTitle": false,
"mimeType": "",
"hasMimeType": false
},
{
"name": "Evaluation",
"content": "For each epoch in the Testing.zip file, participants need to make the corresponding prediction for each epoch according to different usage in the Testing_Info.csv file and submit result_name_date.csv file\n\n* The usage 3 indicates the Identification mode. Participants need to judge whether the epoch is from an intruder. If yes, returns 0. If not, return the predicted subject ID.\n* The usage 4 indicates the verification mode. Participants needs to judge whether the epoch is from subject as he or she claimed. If yes, return 1. If not, return 0.\n\nA baseline method is provided with the following link,\n[Baseline.zip](https://storage.googleapis.com/kaggle-competitions-data/kaggle-v2/37941/4046133/compressed/Baseline.zip?GoogleAccessId=web-data@kaggle-161607.iam.gserviceaccount.com&Expires=1659170547&Signature=xCow0WpxZ%2FJLcy48CSpMBPpRxFLiJtduvJ1NPrv6jo0Uw80OPCAmODe5OfnWOTD2l40cPiIQRAq8A16FPm2GuV%2FuvQtDZ2S%2F6MQNR7IBcbxG9vBFX3xzzE%2Ffw2EarkZmsENkFLQkC%2FNQ9EsD0vBeZqsoI80WWV19u47iOJN5dkrvGBfF7kh%2B2z5VUVzVHjdCIyIHfrgx8dc3SeRs6d1%2B9IosVGpVG2%2B5olkhABafafYjGQI1wf%2BclQQwlnAPX2v7F0Z5ei3uOcTAGzUHPMBhGh5MLTnNwAh38rRD%2BNux%2F5QuvU6p6ZX1D6RS8Q5t%2FNb%2BYBFj%2F%2BbEbzgZ734ahhKCXw%3D%3D&response-content-disposition=attachment%3B+filename%3DBaseline.zip)\n\nThe mean recognition accuracy for all the identification and verification tasks is used as the performance metric for this competiution.. \n\n\n## Submission Format\n\n**For every author in the dataset**, submission files should contain two columns: `EpochID` and `Prediction`. `EpochID` should be the ID of the epoch. `Prediction` shoule be the prediction result for each epoch.\n\n\nThe file should contain a header and have the following format:\n```\nEpochID,Prediction\nepoch057852,43\nepoch057853,43\nepoch057854,1\nepoch057855,6\nepoch057856,0\nepoch057857,0\nepoch057858,0\nepoch057859,52\n```\n\nAn example submission is also provided,\n[Result\\_example\\_0909\\_232857.csv](https://storage.googleapis.com/kagglesdsdata/competitions/38725/4244254/Result_example_0909_232857.csv?GoogleAccessId=web-data@kaggle-161607.iam.gserviceaccount.com&Expires=1663128258&Signature=EiMYqH3wVjkKSVtxlz7DskJJNnq3eYz%2BAO7zm6srrIdmTh8J6JhiP4gtErgUKzYmyeVAi7DLIy6c4JVo%2BWi3TOUI975oO2j0MKPzMDJaFH3xL7bUvpv7pYSgb3lh7UCUQgA%2B%2BY96%2FSNB%2FhXtz0x05uRziqgATJxYCxHlWc8lZ7J8%2Bzx3fYCMci5rr5F1Lvq5xVORTxm0lp1ckivSxh2IwU6OA5ZVWOJwJ9Rra1Gx7A3tzpZTUXi9M7ZijUzyyZNPFlBz90%2Fcr4aNvlQsrNunxrQsQOJU1coL3YdMzLi4IFH3JVzp068Z58PTavtGZKwfSK1BYbMloCIkcBx6ZIuIGA%3D%3D&response-content-disposition=attachment%3B+filename%3DResult_example_0909_232857.csv)\n\n## Reference:\nGan Huang, Zhenxing Hu, Weize Chen, Zhen Liang, Linling Li, Li Zhang, Zhiguo Zhang. M3CV:A Multi-subject, Multi-session, and Multi-task database for EEG-based Biometrics Challenge. doi: [https://doi.org/10.1101/2022.06.28.497624](https://www.biorxiv.org/content/10.1101/2022.06.28.497624v1)[[PDF](https://www.biorxiv.org/content/10.1101/2022.06.28.497624v1.full.pdf)]\n\n",
"isPublished": false,
"hasIsPublished": false,
"postTitle": "",
"hasPostTitle": false,
"mimeType": "",
"hasMimeType": false
},
{
"name": "rules",
"content": "# 1 Qualified Participants\n\n1.\tThis competition is open to the whole society, no age, status, nationality, related fields of individuals, colleges and universities, research institutions, enterprises, start-up teams, etc.\n\n2.\tThose who have access to the questions and dataset in advance are not allowed to participate in the competition. Other employees can participate in the ranking of the competition, but they are not allowed to receive any prizes.\n\n# 2 Requirement\nWe encourage participation by an individual or a team. Each team can have no more than 5 members, and can be formed across universities/institutions/enterprises freely, but each person can only join one team\n\n# 3 Supplement \n\n1.\t**Fair competition:** competitors are not allowed to copy others’ works, exchange answers or use more than one trumpet in the competition. If found guilty, the results will be cancelled and dealt with seriously.\n\n2.\t**Organization statement:** the organizing committee reserves the right to adjust and modify the rules of the competition, the arrangement of the competition, the right to judge and deal with cheating in the competition, and the right to withdraw or refuse awards to the participating teams that affect the organization and fairness.\n\n3.\t**Baseline model:** the baseline model is for the reference of competitors, and can be improved upon. Competitors cannot directly submit the prediction structure of the baseline model.\n\n4.\t**Property:** The entrant agrees to grant the organizer the right to use the results free of charge, worldwide, for an unlimited period of time, including but not limited to the provision of services, further development of services, commercial use and sublicensing. For the avoidance of doubt, the organizer has full intellectual property rights in the new results based on the use of the aforementioned results, and the entrant agrees not to claim any rights or interests in the new results, including but not limited to ownership rights, and to impede the implementation of the new results based on ownership rights in the aforementioned results.\n\n# 4 Anti cheating instructions\n1.\tParticipants are not allowed to register for multiple accounts. Once found, their scores will be cancelled and dealt with seriously.\n\n2.\tParticipants are not allowed to use rules loopholes or technical loopholes and other bad ways to improve their performance ranking outside the scope of technical ability assessment. Once found, their performance will be cancelled and dealt with seriously.\n\n3.\tor those who have access to relevant data of the competition, their submitted works will not be included in the rankings and awards.\n",
"isPublished": false,
"hasIsPublished": false,
"postTitle": "",
"hasPostTitle": false,
"mimeType": "",
"hasMimeType": false
},
{
"name": "Experiment",
"content": "\n\n\nA reliable EEG-based biometric system should be able to withstand changes in an individual's mental state (cross-task test) and still be able to successfully identify an individual after several days (cross-session test). Here we built an EEG dataset M3CV with 106 subjects, two sessions of experiment on different days, and multiple paradigms. Ninety-five of the subjects participated in two sessions of the experiments, separated by more than 6 days. The experiment includes 6 common EEG experimental paradigms including resting state, sensory and cognitive task, and brain-computer interface. To the best of our knowledge, no previous studies have evaluated the test performance of EEG-based biometric algorithms in cross-paradigm and cross-session scenarios on the public dataset with more than 100 subjects.\n\n\n\n1. **Resting-state**: Runs 01 and 14 were resting-state EEG signals with eyes-closed (**EC**), while Runs 02 and 15 were resting-state EEG signals with eyes-open (**EO**), each lasting for 1 min. In the eyes-open run, the subjects were required to keep their eyes fixed in front and blink as little as possible.\n3. **Transient-state sensory**: EEG elicited by visual, auditory, and somatosensory stimuli were recorded in Runs 04 and 11. For each run, 30 trials of **VEP**, **AEP**, and **SEP** were arranged in random order. Each stimulus lasted 50 milliseconds, and the inter-stimulus interval (ISI) was set at 2–4 s. On average, each run lasted 4.5 mins.\n4. **Steady-state sensory**: A train of visual, auditory, and somatosensory stimuli were released in Runs 05, 09, and 12 respectively. Considering the different frequency responses of different modality stimuli, and the higher signal-to-noise ratio of SSVEP compared to SSAEP and SSSEP, the stimulation frequencies and recording times were different. They were 10 Hz and 1 min for **SSVEP** in Run 05, 45.38 Hz and 2 mins for **SSAEP** in Run 09, and 22.04 Hz and 2 mins for **SSSEP** in Run 12.\n5. **P300**:: A visual oddball experiment was arranged in Run 07 with the red square as the **target** stimuli and a white square as the **nontarget** stimuli on the screen. Each square lasted 80 ms with the ISI 200 ms. Six hundred trials of stimuli were arranged in 2 mins, in which target stimuli had a possibility of 5% (30 trials of target stimulation and 570 trials of nontarget stimulation). The subjects were asked to count the number of red squares and report after the run so that their attention would remain on the screen.\n6.\t**Motor Execution**: Executed movement was performed in Runs 03, 06, 10, and 13. During these runs, the subjects were instructed to respond to a visual cue by gripping their left hand (**LH**) or right hand (**RH**), or by lifting their right ankle (**FT**) for a duration of 3 s, i.e., until cue offset. No feedback was provided during the online recording. To minimize the intra- and inter-subject variability, we tried to ensure the consistency of the movements for all subjects. Hence, the executed movement was used instead of motor imagery in classical BCI experiments, and no feedback or training was given to the subjects.\n7. **SSVEP with selective attention**: Six white squares flashed simultaneously at different frequencies in Run 08. First, the target square turned red for 200 ms. After 500 ms, all squares with the frequency of 7 Hz, 8 Hz, 9 Hz, 11 Hz, 13 Hz, and 15 Hz began to flash and lasted 10 s. The subject was asked to focus on the target square during the 10 s. In total, there were 12 trials of **SSVEP-SA** arranged in Run 08. ",
"isPublished": false,
"hasIsPublished": false,
"postTitle": "",
"hasPostTitle": false,
"mimeType": "",
"hasMimeType": false
}
]
}