Metadata collected by sensors

Metadata collected by sensors #

Most acoustic sensors (or passive acoustic monitors) collect some metadata along with audio recordings.

Each vendor typically collects metadata in their own format.

The goal of this document is to provide a common standard for sensor-produced metadata.

However, it is currently unrealistic to expect every vendor to collect metadata in the same format. Even if mass adoption occurs, there are still an uncountable number historically recorded audio recordings for which the metadata could prove valuable.

We expect that the most common pattern will be babel fish pattern - that is, there will be one or more tools that can extract metadata from a recording and convert it to the common format defined in this document.

The Emu tool is one such tool. It currently is in development along with this standard.

Metadata locations #

There are multiple places where metadata for a recording can be found:

  • In the filename of the recording
    • This is the most commonly seen and used metadata
  • In the audio file itself, typically in a text, comments, or custom header
  • In a sidecar file - that is:
    • a separate file stored in the same folder/directory as the audio file
    • in a different format and extension
    • but with the same base name
    • stores data for only the file that it shares its basename with
  • In a shared sidecar file - that is:
    • a file that is not the audio file itself
    • usually stored in a common location, like the root of a memory card
    • in a different format and extension
    • stores data for all files in its directory and below it
  • In any other files, typically unrelated to a recording - that is:
    • a file that is stored in a common location, like the root of a memory card
    • it stores values associated with the sensors deployment, but not explicitly associated with a recording

All known sensors store some metadata in the filename of the recording.

Frontier Labs and Wildlife Acoustics (and others?) store metadata in the audio file itself.

As far as we are aware, no sensor stores metadata in a sidecar files. However, some software (like Raven, and AvianNZ) do store annotations in sidecar files.

Some sensors, like the Frontier Labs BARs used shared sidecar files. Each recording has an entry in a CSV file that includes the metadata in an easy to use format.

TODO: add example of shared sidecar file

Some sensors, like Wildlife Acoustics SM2s store a log of regularly sampled temperature readings in an unrelated file. This is a good example of a metadata file that is not related to any of the recordings.

TODO: add example of shared sidecar file

Recommendations #

The principle here is to make it always easy to move audio files and allow the metadata to be moved along with them, with the least chance of error or erasure.

1. Priority #

Vendors SHOULD store metadata in the formats described above, but, with an emphasis of prioritizing the formats in this order:

  1. In audio files
  2. In filenames
  3. In shared sidecar file
  4. In sidecar files
  5. Other files

2. Redundancy #

Any information stored in the filename of a recording, or in a side car file, SHOULD also be stored inside the audio file itself.

A vendor SHOULD NOT implement on only one metadata format.

The metadata standard #

Data stored in the filename #

  1. Filenames MUST consist of only safe characters.

    • Unsafe characters include the following and MUST NOT be used:
      • Characters that can be used as wildcards in a shell command ( such as *, ?, [, ], {, }).
      • Characters that are used in path constructs (such as /, \, :, and ;)
      • Non-printable characters (\x00 to \x1F)
  2. Filenames SHOULD prefer basic ASCII characters

    • UTF-8 characters MAY be used but there a strong preference for simpler characters
  3. Filenames MUST avoid encoded information

    • Filenames should be readable by humans
  4. Filenames MUST be unique for each recording.

  5. Filenames MUST contain a date stamp

  6. Filenames SHOULD begin with a date stamp

  7. Filenames date stamps MUST adhere to the ISO8601 format modified for filenames.

    • The date stamp SHOULD be in the format YYYYMMDDThhmmss.ssssss
    • The date stamp MAY use the separator - or (space) instead of T but it is not recommended. (This is allowed in the RFC3339 format, but not in the ISO8601 format).
    • The date stamp SHOULD omit the fractional seconds component .ssssss, unless the sensors has a an accurate high-resolution clock
    • If a fractional seconds component is included, the sensor MAY emit only as many fractional digits as are needed to represent the resolution of the sensor’s clock
  8. A filename SHOULD contain a UTC offset when this information is available.

    • Valid formats include:
      • YYYYMMDDThhmmss[.ssssss][+|-]hhmm
      • YYYYMMDDThhmmss[.ssssss]Z
  9. Filenames SHOULD consist of a series underscore (_) delimited fields, suffixed with a file format extension (including a . period).

    <prefix>_<date_stamp>_<suffix>.<extension>
    
  10. Filenames SHOULD NOT

    • use delimiter other than _ to separate fields
    • omit delimiters between fields
    • have leading or trailing delimiters (_a.wav and a_.wav are invalid)
  11. Within a field in a filename, words can be separated by - (a hyphen)

    • a (space) may be used to separate words but it is not recommended
  12. A filename MAY contain either or both of a prefix or suffix around the date stamp

  13. The prefix and suffix MAY contain multiple fields (delimited by _)

  14. A filename MAY contain the GPS location of the file

    • The location MUST be formatted as per the ISO6709 string format (Annex H) but modified for filenames. Effectively this format is:

      [+|-]<latitude in decimal degrees>[+|-]<longitude in decimal degrees>[+|-]<altitude in meters>
      
      • The altitude is optional.
      • The leading + or - is not optional.
      • The trailing / solidus is not acceptable in a filename and is omitted.
      • The coordinates MUST be in the WGS 84 datum.
  15. The filename’s extension MUST match it’s format

Data stored in the audio file #

  1. Metadata stored in headers of audio files MAY use one of the following formats:
    • A custom RIFF chunk in a WAVE file
    • A standard mechanism for storing comments in a FLAC, WAVE, or MP3 file

This standard may define it’s own format for storing metadata in audio files, however, we’re seeking comments from vendors before designing such a format.

  1. Within the metadata section fields are arranged as key-value pairs. Each pair is delimited either by the format’s standard delimiter, or a null character (\0)

  2. Within the key-value pairs, the key is delimited by the format’s key-value delimiter or an = (equals) sign

  3. The key for each key-value pair MUST consist only of the characters a-z, A-Z, 0-9, and _.

  4. The value for each key-value pair MAY by and valid UTF-8 string (format permitting)

  5. The keys present MAY include any of the terms defined later in this document

    • If an analogous field is recorded, the key SHOULD be named the same as is defined in this document.

      i.e. Do NOT include RECORDING_DATE in the key-value pairs when this document defines recording_start

  6. Fields not included in this document MAY be included.

  7. Arrays of values can use a suffixed index to indicate the position of the value

    • These indexes MUST
      • be integers
      • be positive
      • start from 0
    • For example, microphone_type_0 indicates the first microphone type and microphone_type_1 indicates the second microphone type.

Well known fields #

This is a list of the fields that could be defined in the metadata of an audio file:

  • dateTime in ISO8601 format
  • sensorUTCOffset - not required if included in dateTime
  • fileName - ideally, containing dateTime and UTC information
  • recordingStart
  • recordingEnd
  • sensorID - identification of the sensor for deployment purposes
  • deviceSerial - number identifying the sensor to the vendor
  • sensorFirmwareVersion
  • microphoneType_<index>
  • microphoneID_<index>
  • microphoneBildDate_<index>
  • channelGain_<index>
  • sampleRate
  • fileSize
  • duration
  • fileFormat
  • bitsPerSample
  • channelCount - 1 or 2 channels, i.e.: mono or stereo
  • bitSize
  • mediaType
  • memoryCardSerial
  • memoryCardID
  • cardSlotNumber
  • memoryCardCapacity
  • memoryCardFreeSpace
  • memoryCardSpeed
  • memoryCardFormatType
  • memoryCardManufacturerID
  • memoryCardOemID
  • memoryCardProductRevision
  • memoryCardWriteCurrentVmin
  • memoryCardWriteCurrentVmax
  • memoryCardWriteB1Size
  • memoryCardEraseB1Size
  • batteryVoltage
  • batteryPercentage
  • powerType - solar or internal battery
  • location
  • latitude
  • longitude
  • lastTimeSync
  • scheduleName

// more to come