NY Metadata (YAML)

This document describes the standard dictionary structure used in .yml metadata files for EEG time series data in the NY format.

Overview

The YAML format is used to share EEG metadata in both Python and Julia environments. Each .yml file provides comprehensive metadata for its corresponding .npz data file.

Dictionary Structure

The YAML file contains the following top-level keys:

formatversion
acquisition
documentation
id
stim
perf (since formatversion 0.0.3)

`formatversion`

Specifies the version of the YAML metadata format.

Field	Type	Description	Example
`formatversion`	String	Metadata format version number	`"0.0.3"`

`acquisition`

Contains all technical information about the EEG data acquisition process.

Field	Type	Description	Example
`filter`	String	Filter settings of the EEG acquisition machine	`"Low-Pass 83Hz (Butterworth order 4 zero phase) for downsampling"`
`ground`	String	Location of the ground electrode	`"Fpz"` or `"N/A"`
`reference`	String	Location of the reference electrode	`"A1"` or `"N/A"`
`hardware`	String	Commercial name and producer of EEG amplifier	`"g.tec EEG - g.USBamp EEG amplifier"`
`software`	String	Software used for data acquisition	`"OpenViBE, INRIA (France)"` or `"N/A"`
`samplingrate`	Integer	Sampling rate in Hz	`256`
`sensors`	Array of Strings	EEG electrode locations (excluding ground/reference)	`["Fpz", "F7", "F3", "Fz", ...]`
`sensortype`	String	Type and material of electrodes	`"Ag/AgCl Wet electrodes"`

`documentation`

Contains references and documentation links for the dataset.

Field	Type	Description	Example
`description`	String	Link to dataset description	`"https://zenodo.org/records/806023"`
`doi`	String	Digital Object Identifier	`"https://theses.hal.science/tel-01196752"`
`investigators`	String	Principal investigators	`"Alexandre Barachant"`
`place`	String	Institution where experiment was conducted	`"GIPSA-lab..."`
`repository`	String	Link to data repository	`"https://zenodo.org/records/806023"`

`id`

Contains identification information for the specific recording.

Field	Type	Description	Example
`condition`	String	Experimental condition	`"None"` or specific condition
`database`	String	Name of the database	`"AlexMI"`
`paradigm`	String	BCI paradigm type	`"MI"` or `"P300"`
`run`	Integer	Run number within session	`1`
`session`	Integer	Session number	`1`
`subject`	Integer	Subject identifier	`1`
`timestamp`	Integer	Year of data collection	`2012`

`stim`

Contains stimulation and labeling information.

Field	Type	Description	Example
`labels`	Dictionary	Mapping of class names to numeric codes	`{right_hand: 2, feet: 3, rest: 4}`
`nclasses`	Integer	Total number of stimulus classes	`3`
`trials_per_class`	Dictionary	Number of trials available for each class	`{feet: 20, rest: 20, right_hand: 20}`
`offset`	Integer	Offset in samples from stimulation to trial start	`0`
`windowlength`	Integer	Trial duration in samples	`768`

`perf` (formatversion `0.0.3` and above)

Contains classification performance metrics associated with this recording/subject.

For MI paradigm, performances are organized by task pair and classifier.

Field	Type	Description	Example key
`<task>`	Dictionary	Accuracy per classifier for a given task	`right_hand-feet`

Inside each task:

Field	Type	Description	Example
`MDM`	Float	Accuracy of MDM classifier	`0.825`
`ENLR`	Float	Accuracy of ENLR classifier	`0.65`
`SVM`	Float	Accuracy of SVM classifier	`0.675`

For P300 paradigms, perf usually contains global accuracies per classifier (without task subkeys).

Example Structure (formatversion 0.0.3)

acquisition:
  filter: Low-Pass 83Hz (Butterworth order 4 zero phase) for downsampling
  ground: N/A
  hardware: g.tec EEG - g.USBamp EEg amplifier
  reference: N/A
  samplingrate: 256
  sensors:
  - Fpz
  - F7
  - F3
  - Fz
  - F4
  - F8
  - T7
  - C3
  - Cz
  - C4
  - T8
  - P7
  - P3
  - Pz
  - P4
  - P8
  sensortype: Ag/AgCl Wet electrodes
  software: N/A
documentation:
  description: https://zenodo.org/records/806023
  doi: https://theses.hal.science/tel-01196752
  investigators: Alexandre Barachant
  place: 'Laboratoire Electronique et systeme pour la sante CEA-LETI dans l''Ecole Doctorale : EEATS, Universite de Grenoble'
  repository: https://zenodo.org/records/806023
formatversion: 0.0.3
id:
  condition: None
  database: AlexMI
  paradigm: MI
  run: 1
  session: 1
  subject: 1
  timestamp: 2012
stim:
  labels:
    right_hand: 2
    feet: 3
    rest: 4
  nclasses: 3
  trials_per_class:
    right_hand: 20
    feet: 20
    rest: 20
  offset: 0
  windowlength: 768
perf:
  right_hand-feet:
    MDM: 0.825
    ENLR: 0.65
    SVM: 0.675