In-Vehicle Destination Entry By Voice: Practical Aspects

Speech recognition has been shown to increase driver safety in car applications, if the application is well designed. Especially destination entry by voice is not only safer, but also faster and more convenient than the traditional haptic interfaces [1][2]. Building a high performing in-car destination entry system requires tackling a number of practical challenges. The vocabulary is very large, often multilingual in nature, and the application needs to run in the noisy environment of a driving car on hardware with limited resources. Moreover, the type of language drivers want to use is becoming more and more natural, with a higher demand for up-to-date services in a context of other domains such as entertainment being active as well. In this paper, we will go over some practical aspects this all brings to the problem of designing a speech recognition system for destination entry.