1Connected Speech in Cars
The Fully Networked Car Geneva, 3-4 March 2010
Udo HaiberCOO, SVOX
Agenda
Motivation
2
Connected Speech Services
Effects on Users
Effects on Speech Technology
The Fully Networked Car Geneva, 3-4 March 2010
Effects on Speech Technology
Effects on Stakeholders
3
ABOUT SVOX
The Fully Networked Car Geneva, 3-4 March 2010
ABOUT SVOX
SVOX portfolio 4
In-Car Communication
Acoustic Signal Enhancement
ASR Engines
The Fully Networked Car Geneva, 3-4 March 2010
TTS Engines Dialog Engine Integration Framework
5
MOTIVATION
The Fully Networked Car Geneva, 3-4 March 2010
MOTIVATION
Motivation – Always connected generation 6
Video – Always Connected
The Fully Networked Car Geneva, 3-4 March 2010
Source: PEGA Design & Engineering
Motivation – Speech solutions
o Always connected generation wants to use
mobile internet also in cars, BUT...
o Driving safety must not be decreased
7
o Driving safety must not be decreased
Always connected
Driving Cars
Speech
The Fully Networked Car Geneva, 3-4 March 2010
�Speech as hands-free, eyes-free solution
Maciej, J. & Vollrath, M. (2009).
Comparison of manual vs. speech-based interaction with in-
vehicle information systems.
Accident Analysis and Prevention, 41, 924–930
8
CONNECTED SPEECH
The Fully Networked Car Geneva, 3-4 March 2010
CONNECTED SPEECH SERVICES
Todays In-Car Services 9
Communication SpeechInp SpeechOut Connected LBS
Phone / name dialing � �
SMS, eMail �
Driving support
Destination input / directions � �
POI search �
Traffic messages � �
Infotainment / ConvenienceInfotainment / Convenience
Music, Video � �
Safety & Security
eCall � �
CRM
Remote Diagnostics �
The Fully Networked Car Geneva, 3-4 March 2010
Future In-Car Services 10
Communication SpeechInp SpeechOut Connected LBS
Phone / name dialing � �
SMS, eMail � �
Social networks � � � �
Twitter � � �
Driving support
Destination input / directions � �
POI search � � �
Business Listing � � � �
Traffic messages � � �
Floating Car Data � �
Parking � � �
Speech Traps � � �
Eco driving �
Infotainment / Convenience
Music, Video � � �
Travel Guide � � � �
Weather � � �
News, Stocks, Sports � � � �
Wiki � � �
Events � � � �
Shopping � � � �
The Fully Networked Car Geneva, 3-4 March 2010
Shopping � � � �
Calendar � �
Web browsing, searching � � �
Safety & Security
eCall � �
Stolen Vehicle Tracking � �
CRM
Remote Diagnostics �
Vehicle Homepage �
SW-update / App store �
11
EFFECTS ON USERS
The Fully Networked Car Geneva, 3-4 March 2010
EFFECTS ON USERS
Effects on Users
Traditional UsersTraditional UsersAlways connected
GenerationAlways connected
Generation
12
hierarchical browsehierarchical browse
prepare, plan things to doprepare, plan things to do
privacy concernsprivacy concerns
GenerationGeneration
keyword searchkeyword search
more spontaneous, cause everything is available
always and everywhere
more spontaneous, cause everything is available
always and everywhere
expose privacyexpose privacy
The Fully Networked Car Geneva, 3-4 March 2010
privacy concernsprivacy concerns
single-taskingsingle-tasking
expose privacyexpose privacy
multi-taskingmulti-tasking
13
EFFECTS ON SPEECH
The Fully Networked Car Geneva, 3-4 March 2010
EFFECTS ON SPEECH TECHNOLOGY
Effects on Speech Input
Engine User Interface
14
Engine
• Large vocabulary fuzzy matching
• Embedded vs. server follow data => hybrid
• Enrollment vs. Pre-defined vocabulary
User Interface
• Hands-free mode needs DIALOG to present and select from possible answers
• Seamless integration of on-/offboard interaction (e.g. One voice, one concept,...)
• Extensibility
The Fully Networked Car Geneva, 3-4 March 2010
• Extensibility
• Traditional approach as legacy feature
Effects on Speech Output
Prompt text length increases
Prompt text dynamics increases
Prompt text less well-formed
15
increases (e.g. eReader)
• Naturalness must be increased, in order not to bore listerners
• Audio-Streaming of Server TTS
dynamics increases (e.g. RSS feed)
• Pure TTS prompts, no pre-recording (as for turn-by-turn nav) anymore
• Learning TTS (adaptive)
well-formed (e.g. Mail)
• Focus on text pre-processing
• Robust language identification used to handle polyglot texts
• 2D-structures to
The Fully Networked Car Geneva, 3-4 March 2010
• 2D-structures to enable mail, web content
16
EFFECTS ON
The Fully Networked Car Geneva, 3-4 March 2010
EFFECTS ON STAKEHOLDERS
Effects on stakeholders
Commercial Side Technical Side Legal Side
17
• More players, more complexe business models
• Traditional: OEM, Tier1
• Future: OEM, Tier1, Carriers, Handset-OEMs, App Stores,
• More developers (app store) not only Professionals
• need for open software concepts with risk of reduced Quality
• Responsibility for recalls, accidents, etc.
• Liability
The Fully Networked Car Geneva, 3-4 March 2010
App Stores, Content/Service Provider
Quality Assurance
18
SUMMARY
The Fully Networked Car Geneva, 3-4 March 2010
SUMMARY
Summary
Speech solutions exist now for decades, but acceptance will increase remarkably with this new field, because...
Speech is advantageous over traditional UI‘s, when
19
Speech is advantageous over traditional UI‘s, when searching large lists especially in automotive environment
Products showing this advantage will enter the market place already this year
The Fully Networked Car Geneva, 3-4 March 2010
Always connected
Driving Cars
Speech
SVOX – Your Dialog Partner 20
The Fully Networked Car Geneva, 3-4 March 2010
Top Related