[Ltg] HAIL Seminar (11th March): Dr Fang Chen, NICTA

[Andrew Lampert]

                      H.A.I.L. Seminar series
                          CSIRO ICT Centre
                   http://www.ict.csiro.au/HAIL/


Title:          Automatic Cognitive Load Detection Based on Speech Features

Speaker:        Dr Fang Chen
                Senior Principal Researcher/Research Group Manager
                Making Sense of Data Theme
                NICTA

Date:           Tuesday 11th March 2008 at 11am

Location:       CSIRO ICT Centre,
                Building E6B, Macquarie University.

                See <http://www.ict.csiro.au/HAIL/location.htm> for details.

Video:          We usually stream live video of seminars.

                At the seminar time (see above), point your browser at:
                  <http://webcast.nsw.csiro.au/httpfs/ICT/HailSeminar/live.asx>


Abstract

Monitoring cognitive load is important for the prevention of faulty errors in task-critical operations, and the development of adaptive user interfaces, to maintain productivity and efficiency in work performance. Cognitive load variations have been found to impact multimodal behaviour, in particular, features of spoken input. Our research goal is the implicit, objective, automated and real-time estimation of a user's cognitive load based on human's natural speech features, suitable for high-complexity, real-time deployment.

In this talk, we focus on potential speech feature indices. We present our attempt to induce controlled levels of load and solicit natural speech, and the use of machine learning in the development of a speech classifier that is able to detect different levels of load in the speech signal. The user studies confirm a significant variation of speech features in different cognitive load levels.

To capture the variation information of speech features and output the cognitive load level as discrete level ranges, we propose a novel, fully automatic, speaker-independent classification approach to monitor cognitive load level by using speech features. In this approach, a Gaussian Mixture Model (GMM) based classifier is created with unsupervised training. Channel and speaker normalization are deployed for improving robustness. Different delta techniques are investigated for capturing temporal information. And a background model is introduced to reduce the impact of insufficient training data. The proposed automatic cognitive load detection system achieves 71.1% and 77.5% accuracy in two tasks respectively.

To validate this proposed approach in a real-world scenario, we investigate a real life bush fire management situation. In this scenario, incident commander exchange information with and give orders to other roles mainly by speech. The proposed system is tested in this scenario and receives reasonable results considering the much more background noises and lower speech quality. This experiment shows a great potential in real-world applications.


Short resume

Dr. Chen was employed with Beijing Jiaotong University in China from 1995-1999.  She was appointed Associate Professor of the Faculty of Electronic and Information Engineering in 1995, the Deputy Director of the Institute of Information Science in 1996 and then Dean of Faculty of Electronic and Information Engineering in 1997. She began her career in industry in 1999 as senior researcher and Team Leader of Text-to-Speech Group in Intel China Research Centre. She joined Motorola in 2000 as a Principal Researcher, and was the founding research manager of the Speech and Language Generation Research Lab of Motorola China Research Centre, where she also acted as the account manager of business relationships for the Centre.

Dr Chen moved to Australia in 2002 to work for the Motorola Australian Research Centre, where she chaired the Patent and Publication Committees. She jointed NICTA in 2004 and is currently Senior Principal Researcher and ATP research group manager in the theme of making sense of data. Dr.Chen has received visiting Professor and Honorary Associate positions with the University of New South Wales and the University of Sydney.


----------------------------------------
The HAIL Seminars' URL:
http://www.ict.csiro.au/HAIL/

Contacts:       Andrew Lampert
Address:        CSIRO HAIL Seminars,
                c/o Andrew Lampert,
                Locked Bag 17,
                North Ryde NSW 1670

Phone:  (02) 9325 3100
Email:  Andrew.Lampert at csiro.au


*Administration*
-----------------------------------------------------------------
* To leave the list, send the message "unsubscribe HailSeminars" to
the list server address <majordomo at nsw.cmis.csiro.au>.
The subject line of your e-mail message will be ignored.
==================================================================