Signal Processing - November 2016 - 58

Expectation for better in-vehicle CID systems

Multimodal Sensing, Internet of Things

Safety-Optimized Autonomy

advisories, and communicate properly based on their cognitive
and emotional state (Figure 3).
In smarter cars with high autonomy and connectivity, speech
A successful deployment of sophisticated in-vehicle CID
systems need more conversational intelligence, as we have
systems in the future would require a breakseen. Fortunately, the popular demand for
through in the system development process
better IPAs and mobile voice technology is
A key aspect with
from specification, development, testing,
driving improvements in conversational
in-vehicle dialog
and validation in the automotive industry to
agents by increasing conversational intellisystems is that the
ensure high-quality but low-cost software.
gence in many of the ways previously menThe complexity in introducing many additioned in this article to be necessary for
interaction is often
tional sensors into the vehicle combined
in-vehicle CIDs. Integrating developments
carried out while a driver
with much more content and many more
in safety-optimized autonomy and converis operating a vehicle
services from the cyberworld should not be
sational intelligence offers much promise
as a secondary task.
underestimated. It is quite possible that addifor the CIDs that can meet automotive
tional new layers will be introduced in the
requirements for naturalness, ease of use,
infotainment architecture to simplify the development and testlow cognitive demand on users, and minimal distraction from
ing process. New standards may need to be introduced to faciliinteracting with the CID itself. Of course, automobiles constitate industry-wide collaboration and incentivize the adoption of
tute a unique environment for speech in a multimodal setting,
new technologies with affordable cost. Additional features may
and additional research is needed specific to the automotive
be also introduced to support the privacy and security in both
milieu, as discussed previously.
the physical or cyberworlds by using CID systems to recognize
Moving forward, we foresee that CID systems would inand track drivers' identity and set up proper access or operation
creasingly offer explanations about the vehicle itself, including
restrictions while driving.
functional operations, vehicle status, maintenance requirements, or even recommended driving styles for extended uses
or environmental impact. CID systems will further act as a
Conclusions
mediator to synchronize the content and services from different
An in-vehicle dialog system is a complex system that involves
IPAs and integrate them with in-vehicle information. They will
broad interdisciplinary knowledge and technologies from
collaboratively support drivers on various activities with expert
automatic speech recognition, spoken language understanding,

Queries About
Operating Car

Coordination with Driver's
Current Activities, Cognitive,
and Emotional State

Explanations
from Car

Two-Way interactive,
Collaborative, Natural,
Intelligent Communication
Gap: Ecologically Valid Data for Studies of
Behavior and for Computational Models

Harness for
Conversational
Intelligence

Coordination of Conversation
(Addressee Detection, Fluency,
Turn-Taking, and Grounding)
Harness for
In-Vehicle
Environment

Speech Recognition

Annunciation
(Voice Alerts)

Personal Assistants and Mobile Voice Technology
Conversational Intelligence

FIGURE 3. Future directions of CID systems.
58

IEEE SIgnal ProcESSIng MagazInE

|

November 2016

|
Table of Contents for the Digital Edition of Signal Processing - November 2016
