Dr.-Ing. Christian Müller
Senior Researcher Dr. Christian Müller is leading the automotive IUI projects CARMINA, SIM-TD, and various related industry projects. His research focus is user-adaptive multimodal HMI (Human-Machine Interface) with a special focus on speaker classification, which is seen as a source of information for non-intrusive user modeling. Application areas of his research are primarily automotive interfaces and telecommunication applications. From 2006 to 2008, he has been a visiting researcher at the International Computer Science Institute (ICSI) in Berkeley, California. Dr. Müller earned a Ph.D. in Computer Science at Saarland University in January 2006.
His dissertation "Zweistufige kontextsensitive Sprecherklassifikation am Beispiel von Alter und Geschlecht" (Two-layered Context-Sensitive Speaker Classification on the Example of Age and Gender) was supervised by Prof. Wolfgang Wahlster and graded Magna Cum Laude. From 1994 to 2001, he studied Computational Linguistics at Saarland University where he graduated with distinction. He also holds an IT-Business Apprenticeship.
In 2001, he began his research in the area of user-adaptive multimodal system driven by the question how we can design mobile, speech-enabled systems that adapt their behaviors according to the specific needs of a group of users. The first user group he studied were people who are experiencing high cognitive load like for example travelers navigating through a crowded airport trying to find their gates while shopping for last-minute gifts for their families. The next group was the Elderly - and there were two reasons for that: First, access to technology was (and still is) often obstructed for seniors and hence there were efforts to "e-include" them by choosing suitable interaction methods. Second, the elderly represent an ever growing market for IT products that had long enough been neglected by industry. He and his colleagues came up with a mobile shopping assistant and a mobile personal navigator recommending goods respectively paths. However, one issue he had to deal with was: Where does the information come from? How does the system know that the user is under cognitive load or is an elderly person?
On the search for non-obtrusive ways to acquire a user model, he eventually investigated speech as a source of information. As a consequence, for the following years his research derived into the area of signal processing, investigating symptoms of cognitive load in speech and developing an automated method to recognize speaker age (and gender). With fair success: After the first prototype was presented at the CEBIT trade fair in 2005, Deutsche Telekom started to sponsor his research with goal to utilize it for telephone-based product recommendation systems. Then, during his post doctoral time in ICSI's strong speech group (ICSI is the International Computer Science Institute in Berkeley, CA), Christian Müller was able to substantially deepen his knowledge on speech processing as well as machine learning which helped improving his algorithms significantly. As a consequence, his age recognition method was used in the contest of the German Voice Award in 2007. In competition with more than 25 other spoken language systems, the T-Mobile system based on his research results was selected as the winner of the "The Most Innovative German Voice System 2007". Additionally, the method has been patented with the German patent office. Since then, continuous improvement assured the state-of-the-art status of the approach.
Another example of Christian Müller's success in the area of speech-based classification is a research project sponsored by the Volkswagen Electronics Research Laboratories (ERL) in Palo Alto (USA) on acoustic event classification for in-car applications, which he was leading on behalf of ICSI.
Christian Müller edited a book on Speaker Classification that was published in 2007 by Springer in the LNCS/LNAI series and constitutes a state-of-the-art survey of the field. The two-volume book addresses questions such as which speaker characteristics are manifested in voice and speaking behavior, which characteristics can be inferred from analyzing the acoustic realizations, how can this information be used, which methods are most suited to solve problems in this area of research, and how should the quality of the results be evaluated. In parallel to the printed book, the articles are published in full-text electronic form via the Springer internet platform www.springerlink.com. In December 2008, Springer proposed a new edition of the first volume "Fundamentals, Features, and Methods". Unlike the first edition, the second one will be a hard cover book whichis expected to have an even higher impact than the first one. The book will appear in January 2010.
Please refer to the list of publications.