RealTimeSubtitle - 2016/17 - RICM5

Supervisors:

• Didier Donsez
• Jérôme Maisonnasse
• Laurent Besacier

Resource person:

• Marie-Paule Balicco, director of Grenoble disabled person's reception

Members of team:

• David Bui, Communication manager
• Anna Bruel, score coordinator
• Maxime Lechevallier, lead manager
• Tanguy Mathieu, scrum master
• Adrien Mouret, project coordinator
• Abdukmelik Sak, Internship DUT R&T 2nd year
• Andréas Gillet Pascal, Internship DUT R&T 2nd year
• Nicolas Bosia, Internship DUT R&T 2nd year

Objective

• needs by the Grenoble disabled person's reception service
• achievement of an application which semi-automatically record the voice.
• The HMI facilitates the correction of wrong translations thank to collaborative participation

• Who the project is aimed ? All students, particularly, people is hard of hearing, have difficult to note down or have focus problems.
• What ? Students have understanding problems of course.
• Where ? University of Grenoble
• When ? During lecture course mainly. It can be also used during practical or exercice course
• How ? Web platform and web browser
• Why ? Increase the daily life of students allowing the access of education easier.
• How many ? Nearly 60 000 students in Grenoble, whose 850 are persons with disabilities in Grenoble on.

Technical constraint

• recognition API: Dragon/Sphinx/Amazon/Google natif/Google speech
• HTML5 + CSS + Javascript

Logbook

Sprint 1 (January 23th - January 30th)

Objective: define accurately the requirements and the first sprint.

• Understanding of the subject, constraints and available equipments with Jérome Maisonnasse at FabLab. The team hesitates to use a Fablab's microphone or a simplier bluetooth microphone. We are awareness of different technologie (amazone, google natif, pocket sphinx...) and have the possibility to meet researchers ( Michel Vacher and François Portet )
• Determine the customer requirements with Marie-Paule Balicco at the disabled house.
• Brainstorming to define the specifications.
• Setting up the tools to launch the app: meteor lamp
• Setting up the management tools: gitlab, trello, slack.
• Recovery the old project making in RICM4

Sprint until the 30th of January:

• David Bui: is looking for with API more suitable to the acquisition of the voice. There are the choice between Dragon, Spĥinx, Amazone and Google natif
• Anna Bruel: edit the SRS and the view of the HMI.
• Maxime Lechevallier: is looking for the editing part for the collaborative correction
• Tanguy Mathieu: is looking for the integration of pdf in the app.
• Adrien Mouret: is looking for the deployment of server checking portability.

Sprint 2 (January 30th to January 6th)

Conclusion of sprint 1:

• After the meeting with Laurent Besacier and the David's research, we are going to use Google speech API which seems the most efficient. The problem is that the API automatically stop after one minute.
• For the editing part, Maxime looked for a pluggin which calls shareJS. It allows to modify the text like a drive or a text document. This solution is better than the old project and more ergonomic.
• The deployment is done by Adrien on the amazone server at the address IP following: http://35.157.82.44:3000/. The application works well.
• Anna worked on the HMI ergonomy and Didier Donsez introduced us Inspinia which offers packages meteor for make a beautiful HMI. It uses the boostrap technology.
• Tanguy works on the addition of pdf file to reveal. The solution consist of splitting each pages of pdf which will be changed in images, in order to integrate to reveal.

Objective of new sprint 2 until the 6th:

• David Bui: works on the uploading of pdf file.
• Anna Bruel: works on AudioSeg (advised by Laurent Besacier) to split the audio flow following the timer (every 10 seconds) or following blanks.
• Maxime Lechevallier: is looking for the microphone stopping problem, automatically resetting the microphone every 10 seconds.
• Tanguy Mathieu: follows the work on the integration of pdf in the app.
• Adrien Mouret: integrates tracks of audio flow in the app. This tracks should correspond to those recovered with AudioSeg.

Week 3 (February 6th to February 13th)

Conclusion of sprint 2:

• After the meeting with Renaud Blanch, for the edition of text, we keep the idea of the old project with a score system. We don't use a system like google doc. Therefore, we are going to improve that it was doing. A box for each words and don't move when we are over the word. A symbol to add or to remove could be displayed. A coloration system to know if the word has been modified or deleted.
• Finally, Google speech is not free. We use the web speech api. The micro is always stopped after one or two minutes, because there is a problem with the state machine.
• The uploading file is not working yet.
• AudioSeg is not a good solution, it's better to improve the web speech api.
• The integration of pdf is always searching.

Objective of new sprint 2 until the 13th:

• David Bui: sets up the edition of subtitle following the Renaud Blanch's advices.
• Anna Bruel: works on the microphone stop. Tries to improve web speech api.
• Maxime Lechevallier: do pair programming helping Anna Bruel and Tanguy Matthieu simultaneously.
• Tanguy Mathieu: follows his work.
• Adrien Mouret: sets up an https to secure the app.

Week 8 (March 13th)

soutenance