Speak To Me! - DaimlerChrysler's New Speech Assisted Navigation System

• Convenient destination input for navigation

• Enhanced safety through reduced driver distraction

STUTTGART - November 8, 2006: DaimlerChrysler’s researchers are developing a rapid, easy-to-operate speech input system for specifying navigation destinations. “Information refinement” is the name of this new method, which is oriented toward human speech and makes dialogue with the navigation system as natural as possible. The challenge of ambiguities

There are more than 74,000 cities and 970,000 streets in Germany alone. Of these, more than 58,000 city names and 338,000 street names require to be differentiated, since countless opportunities for confusion exist: there are 29 different towns by the name of “Neustadt”, for instance, and countless streets called “Hauptstraße”; the speech recognition system must also be able to distinguish between similarly sounding words such as “Berlin” and “Bellin”.

Compacted knowledge

Pinpointing a specific destination often calls for additional data. The DaimlerChrysler researchers have determined the most suitable forms of information by measuring the average level of drivers’ knowledge. Their study revealed that users usually knew a nearby large town or the relevant federal state, region or river. On the other hand, most users were unfamiliar with license plate designations, zip codes and telephone dialing codes.

The intelligent strategy for the clarification of destinations is known as “Information refinement”. The number of possible destinations – the volume of calculated hypotheses – is reduced step by step. The driver’s tested level of knowledge is taken into account here. For practical application, this means that the essential components (city and street) are nominated first of all. Ideally, these items will already make for mutual clarification; only if the information is still ambiguous are additional optional questions asked (nearby city etc.). For example, if the driver nominates “Frankfurt, Hauptstraße”, the system will ask him or her to name a nearby river. If “Oder” is given, the driver is then only asked for definitive confirmation of the destination “Frankfurt/Oder, Hauptstraße”; this reduces the number of steps in the dialogue.

What additional information the computer asks for is determined on the one hand by the expected level of knowledge of the driver; the other component consists in the various approaches to resolving any ambiguities. The new system can thus operate as conveniently and effectively as possible.

Tried and tested

The scientists have already tested this new speech input system in the laboratory. They selected male and female test persons who had only limited practice in the operation of navigation systems and only little experience of speech input. However, DaimlerChrysler’s new method was compared not only with a conventional speech input system, but also with a manual input method.

In order to determine the respective differences in ease of operation and thus in driving safety, the test persons were required to drive a car while speaking to the computer. In the simulator, measurements were carried out as to how precisely the drivers kept on course; the average deviation from the midline served as an objective measure of driver distraction. The result was greatly reduced distraction while driving, and thus greater safety. The researchers also asked for subjective data – in other words the users’ own impressions. The outcome was a clear preference for speech input, since it outperforms manual input in terms of simplicity, efficiency and rapidity.

Compared with speech input systems used to date, the researchers have reduced the number of dialogue steps by ten percent and the incidence of unsuccessful dialogues by half. The driver must perform a minimum of two steps – naming the city and street, and then confirming this information. Even in difficult, ambiguous cases the dialogues are limited to six steps at the most. The aim of the DaimlerChrysler researchers is to ensure even shorter response times in future and to facilitate natural dialogue.