Natural Language and Voice Control in Automotive Systems

							A Voxi Business Case:


      Natural Language and Voice
      Control in Automotive Systems
      VOICE CONTROL IS HIGH ON THE WISH LIST for                       the ability to exactly express the speaker’s in-
      many drivers, faced with a plethora of controls,                 tentions. Instead of constraining the user to
      displays and dials. Today, a wide range of com-                  limited dialog flows or menu like interaction,
      panies within the automotive industry are                        the user has the full initiative and can always
      investigating, developing and selling technology                 give new natural language commands.
      for voice control. The vision is to improve                          This type of interaction is well suited for a
      safety and comfort by letting the driver use the                 car driver, since the main attention can be
      hands and eyes undisturbed for guiding the                       given the guiding of the vehicle. With Voxi’s
      vehicle in traffic, and use the voice to control                 natural language understanding, the inter-
      the ever-increasing set of equipment in the car.                 action flow can freely jump between different
          However, the full impact of voice-controlled                 commands and applications. The driver does
      automotive equipment has yet to be seen,                         not have to pay attention to following a certain
      partially due to the lack of natural language                    dialog flow.
      understanding in mobile systems. Furthermore,
      there are many practical issues that must be                     Adaptation to Different Languages
      overcome to enable the user to talk naturally to                 The automotive market is international with a
      the things around her.                                           high demand on localization. Thus, it is impor-
          Voxi’s Intelligent Speech Interfaces™                        tant to provide a way of adapting an existing
      (ISI™) platform addresses these problems in a                    application to a new language, without having
      unique way. ISI™ is a general-purpose platform                   to make a complete redesign.
      and is useful in many environments, ranging                          To achieve this, the ISI™ platform makes a
      from automotive systems and mobile systems in                    clean separation between language-dependent
      general to telephony services and consumer                       information (vocabulary and grammar), and
      electronics.                                                     language-independent information (application
          This paper discusses some key issues, and                    model and the related concepts). An appli-
      shows how they are addressed with the ISI™                       cation can quickly be adapted to a new
      platform.                                                        language by just changing the vocabulary and
                                                                       grammar modules. The application logics and
      Natural User-controlled Dialogs                                  interaction design can be reused with no or little
      The ISI™ platform from Voxi focuses on the                       modification.
      unique features of speech: its conciseness and

                  Voxi’s solution enables a natural dialog with the devices in the car

                   Play some dramatic music!

                                              Ok, playing “Also Sprach Zarathustra”!

                  Driver: Check my mail!
                  Car: You have one new message, should I read it?
        Driver    Driver: Yes!
                  Car: From Steve, today. Hi, how are you?…
                  Driver (interrupts): Turn on the air conditioning!                              Phone, radio, AC, …
                  Car: Air conditioning turned on!                        Speech-enabled car
                  Driver: Continue reading!
                                                     Rapid Development using Voximizer™
                                                     Using Voximizer™, the ISI development tool,
                                                     applications can be developed using a quick
                                                     development and test cycle. The vocabulary
                                                     and function bindings of an application can be
                                                     changed runtime and thus immediately tested.

                                                     Connections to Other Technologies
                                                     The ISI™ platform has two major modules, the
                                                     speech recognizer and the natural language
                Platform overview
                                                     engine. It is possible to use a speech recognizer
                                                     from another provider in combination with
Control of Multiple Applications                     Voxi’s natural language engine. The system can
The separation of the application model from         even further be customized with special solutions
the vocabulary also makes it easy to provide a       for speech synthesis, speech recognition, etc.
unified interface to several applications. This          This makes it possible to adapt the solution
makes it possible to dynamically collect the voice   to varying demands that may arise, such as
interaction for many different sub-systems in        speaker identification, noisy environments, and
one central unit, and thus enables transparent       integration with the surrounding systems in the
voice control of the different systems found in      car.
a car.
                                                     Summary
User-independence                                    The Intelligent Speech Interfaces™ platform
Voxi’s speech recognizer requires no user adap-      from Voxi has several important features
tation or specific training of the type often        making it suitable for automotive use:
found in high-end dictation systems or low-end           It promotes a user-controlled interaction
voice markup methods. It can instantly be used       using natural language, and it provides a unified
by the end customer without any tedious con-         linguistic interface to multiple applications.
figuration or recognizer training.                       Applications are easy to develop and to
                                                     move between different languages, using the
Flexible Integration in Embedded Systems             Voximizer™ development environment.
The ISI™ platform can be configured to have a            It can easily be adapted and integrated with
small footprint, by tuning several factors.          different embedded CPU solutions, and is
Among the more important factors are the sizes       designed for limited memory and CPU usage.
of the vocabulary and grammar, and how com-
plex the application interaction should be.
    Furthermore, the software is designed to be
                                                      For more information, please contact Voxi:
easy to port between different architectures. If
needed, it can be custom-compiled for specific        WWW:            www.voxi.com
embedded CPU:s, and thus integrated into              E-mail:         info@voxi.com
existing hardware platforms.                          Telephone:      +46 8 453 9050