EUROPEC recording tool

EAGLES EUROPEC recording tool V1.MAY95SLWG--Handbook 18 May 1995 (aptool2.v1 input by Tib

EUROPEC

: [EUROpean Programme d'Enregistrement de Corpus]

License

: non commercial

Author

: J. Zeiliger

Task

:

• Record and playback
• Protocol-driven recording session.
• Visual and acoustic prompts.
• Start and endpointing of utterances
• Signal and label files production

Description

: This is a corpus prompting/recording software making easier the recording of speech databases. It ensures constant recording conditions for various speakers and/or corpus. The recording session is driven by a protocol file (user-made) which describes the way items (components of the corpus) will be displayed on the screen.

Components

:

• ``Standard SAM format'' files production.
• One or two-channels recording and listening facilities.
• Three recording modes (manual, automatic, continuous).
• Parameter adjustment (sampling rate, input and output gain, etc.).
• Sound levelmeter.
• Free / operator controlled session.
• Error treatment.
• Acoustic prompt.
• Easy data verification.
• Safeguarding files.
• Easy data retrieval and manipulation.
• Simplified session.
• User's Guide and many example files.
• User-customizable in any language.

Modularity

:

Extensibility

:

Size

: Version 4.11: 170 kbytes.
( data: examples of speaker, corpus, protocol and label files provided.)

Implementation

: Microsoft C (v5.1)
TMS code for the I/O board

Platform

: SESAM workstation (PC-DOS based), with OROS AU21/22 I/O board.

Languages

: All.

Retargetability

:

Orthography

: Labelling via ASCII.

Examples

: The program is used in laboratories all over Europe.

Status

:

• robust (used in over 30 labs).
• stable
• Windows release under consideration.

Documentation

: User's guide (30 Pages) SAM-ICP-O45 , March 31st, 1991

Upgrades

: No foreseen upgrade. Windows implementation under consideration.

SourceCode

: Provided to members of ESPRIT projects only

Consulting

:

Format

: 1.4 Mb HD floppy disk

Price

: One floppy and mailing costs.

Restrictions

:

• Research purpose only.
• Citation

Contact

:

Address

: Institut de la Communication Parlee, Grenoble, France.

The EUROPEC prompting/recording tool has been clearly designed to make easier the recordings in large recording campaigns for the constitution of speech databases. However, it provides also simple and basic functionalities. Some particular points must be developed here:

Data authentification

The EUROPEC software ensures automatic and consistant filenaming of the files it produces. The uniqueness of a filename during a recording campaign is ensured by an incremental serial number. For each recording session, the components: speaker, corpus to be recorded, recording conditions, protocol to be used, have to be registered in the relevant DB files.

Protocol of corpus

The protocol-based approach uses a protocol file that contains instructions to be executed by the program during the recording session. These instructions address the prompting of items on the screen and the design of the session. They define how and when items will be displayed, and allow training phases before the recordings. Here are some examples of the relevant codes implemented:

• CLS: CLearScreen and put cursor to 0,0
• VON: Video attribute ON: center, blink, inverse, underline, bright...
• VOF: Video attribute OFF
• BEL: ring bell
• MSG: message to display according to valid video attributes
• NWL: newline on screen
• PAU: n (integer) = duration of the pause in seconds
• TRN: set training mode, no recording
• CMT: comment
• GET: n (integer)
  n: number of items to read in the corpus text file, then to display and record (if recording mode is activated) or not (if training mode is activated)
• LOP: -n, m loop in the protocol file
  n: number of one-line-back in the present file m: number of iterations
• REC: recording mode activation
• RET: waiting for a RC (return)

In a research environment a prompting/recording tool can easily be extended or retargeted with usefull functionnalities to cover other experiments and researchers needs. For instance, local variants of the EUROPEC software have been successfully produced to perform adequate experiments in the following domains:

Perception:

The acoustic prompt can be used to perform perception tests. Stimuli are played to the listener that is then presented menu options to indicate his answer. The answers are stored in the label file (results of the test).

Prosody:

On-line listening facility is provided when the recordings aim at matching a particular ('spontaneous') prosodic profile. The session is a sequence of record, listen/accept/reject (menu options), record, listen/.. phases.

Acoustic:

The acoustic prompt can be used to play blank noise during articulatory positioning then simultaneous I/O is used to record the production of the vocal tract excitated by a blank noise.

All these examples demonstrate that prompting, recording and playback facilities can appear in different ways in various experiments. The protocol-based approach as initiated in EUROPEC is adapted to the definition of the experiment session by the user. This approach must be clearly enhanced: a recording tool should be better considered as a software platform providing basic functionnalities as signal input/output, acoustic or text prompts, storage of labels or results, mouse driven menu. The modularity of these functions would allow the user to design his experiment and write the corresponding protocol file in a high level language; the protocol being then interpreted by the program all along the session.

)