In-vehicle navigation systems are becoming increasingly common. Significant deployment is occurring in Japan, with reported 20% new-car penetration. In Europe, the number of navigation systems sold has doubled every year for the last several years. Likewise, the number of (car) models newly offering navigation systems has doubled every few years, with over 70 models offering navigation systems in CY 2000. North American sales lag European sales, but are expected to overtake European sales by the middle of this decade. In-vehicle navigation systems offer many benefits to the driver. But, since most systems on the market today can be used while the car is moving, they are a source of visual and cognitive distraction. One-way driver distraction can be reduced and satisfaction increased is to provide a voice interface to the navigation system. However, the complexities of a noisy vehicle cabin are compounded by the resource limitations of embedded in-vehicle navigation systems and the tens of thousands of place, street, and other names in a typical in-vehicle database. To address this problem, we have surveyed the capabilities of embedded ASR and TTS engines, identified various ways that people in North America and Europe specify a destination, interviewed car companies and automotive suppliers to understand their in-vehicle speech requirements, and performed statistical surveys of the NAVTECH database to understand issues and approaches to expanding ASR and TTS applications in the car cabin, with specific reference to use with in-vehicle navigation systems. This paper discusses issues in integrating voice engines with in-vehicle navigation systems from a navigation database perspective, summarizes our findings related to destination selection in the car, and outlines steps required to successfully integrate voice engines and navigation systems.
Abstract