Find Speech Trax in Ina's Research area.
Speech Trax is a navigation interface in Ina's TV and radio collections that is based on multiple speech processing techniques.
The main purpose of putting this demo online is to demonstrate the latest results, to highlight the benefits of speech processing techniques in helping to annotate multimedia content and to propose new ways of browsing archive contents.
Results of speech/music discrimination, speaker diarization, speaker retrieval and automatic speech transcription are displayed in the Speech Trax demo.
Speech and speakers carry a heavy part of what makes a program understandable for a human being. Therefore it is of great interest to study how the speech load is shared.
More than 250 hours of radio and tv broadcast news and magazines of French public channels (namely France 2, France 5, France 24, France Inter, France Culture, France Info) of March 2014 were processed. In these programs more than 500 unique speakers were retrieved, mostly anchormen, presenters, sportsmen, politicians, experts, etc. To ease the browsing and the visualization these programs were cut in 15 minute sections.
Speech Trax enables a user to navigate inside a corpus of radio and video documents according to the interventions of famous speakers.
For this the user can enter the name of a person he wants to find in the search bar. An autocompletion module will show him if this speaker has been identified in the corpus. Then, the user just needs to select the excerpt of his choice to listen to the speaker's intervention. For TV, the images of the excerpts are automatically extracted based on the speaker's interventions.
Once a program is shown on screen, the user can also browse through the media and access to all the interventions of an identified speaker by clicking on the magnifying glass next to his name.
Speech processing techniques first require the identification of speech tracks. For this, a speech/music discrimination was carried out using a method by the Institut de Recherche en Informatique de Toulouse (IRIT). Adds were manually discarded in order not to deteriorate the results of the speaker segmentation and speaker retrieval steps.
With speech tracks identified, a speaker segmentation has been realised using the LIUM Speaker Diarization software by the Laboratoire d'Informatique de l'Université du Maine (LIUM). Then, the speaker retrieval phase was led based on the LIA-RAL tool of the Laboratoire d'Informatique d'Avignon (LIA).
Speaker retrieval requires the use of voice models computed beforehand. Each model is constructed by gathering several samples of speech of the considered speaker, ideally in various acoustic conditions. More than 1700 voice models were built in this experiment.
Finally, the speech transcript is produced using the Vocapia Research's VoxSigma software.
This work could not have been realised without the help of a lot of people. We would like to thank François Salmon, Jérémy Andriamakaoly and Jim Uro who helped designing the system during their respective internships. We also would like to thank Elisabeth Chapalain for the manual validation of voice segments that allowed the creation of voice models. Finally we would like to thank our academic partners: Sylvain Meignier, Anthony Larcher (LIUM) and Julien Pinquier (IRIT) for their help and advice.
Start demonstrationSpeech Trax is a demo being developed by Ina's Research department. It is currently a beta version. As such, we would appreciate any opinions and remarks you may want to share with us. Whether you are enthusiastic or disappointed with the results, whether you wish to work with us or merely find out more, contact us!
Start demonstrationUnable to load the latest news.
Please fill in the form below
(Last updated 15th January 2015)
When you access the "Research Area" web site at http://recherche.ina.fr/eng and web sites presenting Prototypes (hereafter termed "the Sites"), published and put online by the Institut national de l’audiovisuel (hereafter "INA"), you agree to these general terms of use ("GTU").
The applicable GTU are accessible online on the Sites when you access them.
The "last updated" date is specified at the top of the document.
INA reserves the right to modify these GTU without notice at any time, inter alia to reflect any statutory, regulatory, editorial and/or technical changes. Modifications to the GTU come into effect and are binding upon web users as soon as they are published on the Sites.
By continuing to browse you agree unreservedly to abide by the current GTU and French law.
If you wish to use the DigInPix prototype, you should also accept the Supplementary Conditions of the DigInPix Service.
The Sites, prototypes ("the Prototypes") presented on the Sites (such as Diginpix, Syncnotes, Bobino, Saphir) and content ("the Content") on the Sites (such as audiovisual and audio content, texts, logos, databases, software) are protected by law, more particularly the provisions of the French Intellectual Property Code.
All rights to the Sites, Prototypes and Content are thus strictly reserved.
You do not acquire any intellectual property rights to these elements.
You may only use them under the conditions stipulated in these GTU.
Apart amalia.js video player subject to its own terms of use (GPL v3 license), INA authorizes access to the Sites, Prototypes and Content solely for strictly personal use.
Any use for other purposes, commercial purposes in particular, is subject to INA's prior written consent. (Contact us by e-mail at espace-recherche@ina.fr)
The name and e-mail address you enter on the "CONTACT" form are used solely to send the corresponding message and are not stored by INA.
A cookie is a small piece of data sent from a web site and stored in a user's web browser. It contains details such as the name of the server, an identifier in the form of a unique number, and possibly an expiry date. Servers access them to read and record information.
INA uses cookies automatically whenever you visit the Sites to ensure they are technically serviceable.
For web analytics purposes, INA also uses the Xiti cookie:
Your browser automatically sends us certain standard data, such as your IP address, the type of browser you use, its functions, your operating system, any third-party site from which you may have reached these Sites and the dates and times at which you accessed the Sites.
This enables us to keep statistics on Site visits, measure the general activity of the Sites, pinpoint any errors and ensure that the Sites have adequate material resources. This collected data, in particular your IP address, does not compromise your anonymity.
You are free to accept or refuse this Xiti cookie.
Just change certain settings in your browser to refuse cookies or be notified if a cookie is saved on your computer.
You can also delete cookies stored on your computer to permanently remove the data they contain.
To do so, you can configure your browser as follows:
To delete cookies in Internet Explorer:
To block cookies:
You assume total responsibility and risk for your use of the Sites.
You acknowledge and accept the characteristics, technical performance, limits and risks of the Internet.
You are responsible for taking all appropriate measures to protect your data and/or software and/or hardware against any infection, misappropriation, hacking, viruses, malware or intrusion attempts.
You assume full liability for your use of the Sites, Prototypes and/or Content and for any misuse thereof, in particular unlawful, improper and/or unauthorized use.
The Sites may be interrupted by INA at any time for maintenance purposes or due to cases of force majeure, INA accepting no liability in that respect.
It is agreed that the Sites and Prototypes are supplied "as is".
INA cannot guarantee that the Sites and Prototypes are free from any errors or that any such errors will be corrected. INA provides no personal assistance or hot line.
INA cannot guarantee the accuracy, precision or exhaustiveness of the information provided on the Sites. Accordingly, INA accepts no liability for any inaccuracy or omission therein.
Moreover, INA shall on no account be liable for the use of Prototypes or interpretation of information provided on the Sites.
Generally speaking, INA can on no account be held liable for any direct or consequential damage resulting from connections, interruptions, use or any malfunctions of the Sites or resulting from the impossibility of using the Sites.
These GTU are governed by French law. Any dispute in connection with the applicability, interpretation or execution of these GTU shall be referred to the competent French courts.
These supplementary conditions of use ("Supplementary Conditions") govern the use of your images within the framework of the DigInPix Service proposed by INA on http://diginpix.ina.fr/en.
They supplement the aforementioned General Terms of Use ("GTU").
The fact of uploading images on diginpix.ina.fr Site implies your prior, express and unqualified acceptance of these Supplementary Conditions and of the GTU which are not incompatible with the former.
The applicable Supplementary Conditions are those published on the diginpix.ina.fr Site on the day you use the Service. The "last updated" date is specified at the top of the GTU.
INA reserves the right to modify these Supplementary Conditions without notice at any time, inter alia to reflect any statutory, regulatory, editorial and/or technical changes. Modifications to the Supplementary Conditions come into effect and are binding upon Users as soon as they are published on the diginpix.ina.fr Site.
If you represent a legal entity, you guarantee that you are vested with the necessary authority and powers to accept these Supplementary Conditions.
If you do not accept or are not empowered to accept these Supplementary Conditions, do not submit your images to the Service and hence do not use the Service.
The DigInPix Service offered and operated by INA enables individuals or legal entities to identify visual objects in their images for testing purposes.
DigInPix Service or Service = automatic identification of visual objects in multimedia content (images or videos)
diginpix.ina.fr Site = the web site at http://diginpix.ina.fr/en
Visual object = any nameable entity
Image(s) = image(s) submitted by the user to the diginpix.ina.fr Site for analysis
User = user of the Service and owner or licensee of one or more Images put online for the purposes stated in these Supplementary Conditions
Dictionary = all visual objects grouped by theme
Results = for each Image, the Service gives the User from 0 to several identifications, each one comprising :
The Service is accessible to any natural or legal person who has unconditionally accepted the GTU and these Supplementary Conditions of Use and has the authority to do so.
The Service is provided to the free of charge, for non-commercial and non-exclusive use for the purpose of experimenting with the Service. The User therefore undertakes to use the Service solely for the purpose of identifying visual objects in its Images.
Any use for other purposes, commercial purposes in particular, is subject to INA's prior written consent (Contact us by e-mail at diginpix@ina.fr).
Any automated use of the Service is prohibited.
You shall not :
The Service is not designed for storing Images; it is up to the User to personally take the necessary steps in this respect.
You retain all your intellectual property rights to the Images you submit to the diginpix.ina.fr Site.
You are solely responsible for the Images you submit to the diginpix.ina.fr Site, as INA can on no account be responsible for Users' Images.
By submitting your Images to the diginpix.ina.fr Site, you guarantee INA that:
You undertake to release INA from any liability and guarantee it against any claim, complaint, lawsuit or legal action resulting from or relating to your Images, your use of the Service or further to your breach of the GTU and these Supplementary Conditions. You undertake to pay INA compensation for any resulting damages, claims, costs and expenses (including lawyer's fees).
You agree that INA may use your Images free of charge and non-exclusively, as part of the Service accessible worldwide, for the following purposes:
In this respect, you agree to INA storing your Images and automatically analyses by means of the Service.
Within the functional framework of the Service, the Image(s) and Results of the automatic analysis are made available to the User by means of a URL that INA sends only to the User.
The User alone is responsible for disclosing the URL to any third parties of its choosing for the purpose of sharing the Images and Results; INA shall on no account be liable for the use of the URL by the said third parties.
INA may on no account disclose, assign, transfer or distribute the Images to third parties, nor use them for any purpose other than the Service.
INA reserves the option to delete your Images, Results and related URLs at any time, without notice or any form of compensation.
You may ask INA at any time to delete your Images, Results and related URLs; to that end, use the "Contact" form or send an e-mail to diginpix@ina.fr stating precisely which URLs you wish to have deleted.
Without prejudice to the User's rights to the Images, the Image-related Results generated by the Service belong to INA.
You may not copy, modify, sell or distribute them to third parties, nor may you create works or derivative products based on the Results.
You may not grant access to or distribute them to third parties under conditions other than those expressly permitted by these Supplementary Conditions.
As hoster of Images, and in accordance with its statutory requirements, INA retains Users' identification data (inter alia the IP address).
This data is processed to comply with the provisions of decree no. 2011-29 of 25 February 2011 on the retention and communication of data identifying anyone who has contributed to the creation of online content. INA alone receives this data.
As required by the "Informatique et Libertés" (French data protection) law of 6 January 1978 modified in 2004, you have the right to access and correct the data relating to you. You can exercise this right by writing to CIL, Legal Department, INA, 4, avenue de l’Europe, 94366 Bry-sur-Marne Cedex France, or by sending an e-mail to cil@ina.fr.
You may also on legitimate grounds object to the processing of the data relating to you.
You acknowledge and accept the characteristics, technical performance, limits and risks of the Internet.
You alone are responsible for:
As hoster, INA is under no general obligation to monitor the Images.
However, pursuant to current regulations, INA may have to :
To that end, each Image includes a link labeled "Report this Content".
It is agreed that the Service is provided "as is".
INA can on no account be held liable for any direct or consequential damage resulting from connections, interruptions, use or any malfunctions of the Sites or resulting from the impossibility of using the Sites (inter alia any loss, damage, hacking or misappropriation of data and/or Images).
INA cannot guarantee the availability of, or regularity of access to, the Service. You are advised that the Service may be interrupted among other reasons for maintenance purposes, due to failures, technical hitches caused by the communications networks used, malevolent action or any other harm or damage caused to INA's hardware or software.
INA cannot guarantee that the Service is free from any computing errors or that any such errors will be corrected. INA provides no personal assistance or hot line.
Accordingly, INA shall on no account be liable for any interruption in or malfunctioning of the Service for whatever reason, or for the duration or frequency thereof.
INA does not guarantee the reliability, accuracy, precision or completeness of the Results obtained through this Service. Accordingly, INA accepts no liability for any imprecision, inaccuracy, error or omission in or relating to the Results.
Moreover, INA is on no account responsible for the use of interpretation of the Results. Consequently, INA accepts no liability for any use of the Results that is in breach of these General Terms of Use.
INA reserves the right, at any time and without compensation or notice, to close down or suspend the Service. Notification of closure of the Service will be notified on the diginpix.ina.fr Site within a reasonable time.
These Supplementary Conditions are governed by French law. Any dispute in connection with the applicability, interpretation or execution of these Supplementary Conditions shall be referred to the competent French courts.
Answers to FAQs on this site
PUBLISHER
Ina – Institut national de l’audiovisuel
Etablissement Public à Caractère Industriel et Commercial (EPIC, a public undertaking)
RCS Créteil 302 421 193 B
4 avenue de l’Europe - 94366 Bry-sur-Marne cedex
Tél : +33 1 49 83 21 01
PUBLICATION DIRECTOR
Agnès Saal - Chairman & CEO
HOSTING
Ina – Institut national de l’audiovisuel
4 avenue de l’Europe - 94366 Bry-sur-Marne cedex
Tél : +33 1 49 83 21 01
Copyright © - Institut national de l'audiovisuel - 2015