1
Speech Technologies in Cars
and the Role of ITU-T
H. W. Gierlich
HEAD acoustics GmbH
Chairman of ITU-T FG CarCom
The Fully Networked Car
Geneva, 3-4 March 2010
Why Speech Technologies
The driving task
mostly occupied:
visual system
not involved:
talking
mainly involved:
hands and legs
involved:
auditory system
=> Auditory Channel of the human system available
The Fully Networked Car
Geneva, 3-4 March 2010
2
Speech applications
o The main speech applications:
• Speech recognition systems
• Speech dialog systems
• Text to speech systems
• Speech enhancement for communication
systems
• Hands-free communication
• Enhanced in-car communication systems
between passengers
The Fully Networked Car
Geneva, 3-4 March 2010
3
Human interaction
1m
Human conversation:
“orthotelefonic reference position”
The Fully Networked Car
Geneva, 3-4 March 2010
4
Speech dialog systems
1m
Human – machine communication
The Fully Networked Car
Geneva, 3-4 March 2010
5
General Requirements for Human-Machine Communication
o Seamless man-machine interaction requires:
• Superior speech recognition
• Superior speech synthesis
• High quality text to speech systems
• Superior dialog systems
The Fully Networked Car
Geneva, 3-4 March 2010
6
Hands-Free Communication
1m
Bluetooth
FRE
Mobile Nw.
IP…
BSS MSC
The Fully Networked Car
Geneva, 3-4 March 2010
PSTN
DSL
7
General Requirements for Hands-Free Communication in Cars
o Seamless human interaction requires low
distraction form the driving task:
• Superior speech sound quality (in the car and
from car to landline)
• Superior noise cancellation
• Low delay transmission
• Wideband speech is highly preferred
The Fully Networked Car
Geneva, 3-4 March 2010
8
Why Wideband in Cars?
o Wideband services in mobile networks available soon
->
o Enabling wideband telephony (100 Hz- 8 kHz) in cars
• Fullband
• Narrow band (car)
• Wideband (car)
o Efficient use of the high quality audio systems in cars:
•
•
•
•
Getting superior sound quality
Increasing speech intelligibility
Increasing naturalness of a conversation
Reduce drivers distraction due to poor speech quality
The Fully Networked Car
Geneva, 3-4 March 2010
9
In-Car Communication
1m
The Fully Networked Car
Geneva, 3-4 March 2010
10
General Requirements for In-Car Communication
o Seamless human interaction requires:
• Increased intelligibility, esp. from front to back
passengers
• In-Car communication system support not
audible for people in the car
• No artifact under any operation condition
• Adaptive to different noise/driving situations
The Fully Networked Car
Geneva, 3-4 March 2010
11
The Role of ITU-T
12
Definition and qualification
of speech signal processing
Test
Methods
Optimization
of Devices
ITU-T
SG 12 & 16
Speech Dialog
Systems
Speech Terminal
Testing
ITU-T Focus Goup
CarCom
The Fully Networked Car
Geneva, 3-4 March 2010
The ITU-T Focus Group CarCom
Parent study group
ITU-T SG12
Car
Industry
FG-Group
CarCOM
Telecommunication
industry
Universities
The Fully Networked Car
Geneva, 3-4 March 2010
Suppliers
to the
car industry
Algorithm
developers
13
ITU-T: Speech Dialog Systems
In car systems:
• Control of car information systems (telephony,
navigation, car specific functions, …)
Network based systems:
• Control of network accessible functions (telephony,
network based navigation, web-browsing….)
Standardization activities in ITU:
• P.851: Subj. evaluation of dialog systems
• Suppl. 24 to P. Rec.: Parameters describing the
interaction with spoken dialog systems
The Fully Networked Car
Geneva, 3-4 March 2010
14
ITU-T FG CarCOM: Speech Recognition
In car systems:
• Control of car information systems (telephony,
navigation, car specific functions, …)
Network based systems:
• Control of network accessible functions (telephony,
network based navigation, web-browsing….)
Standardization activities in ITU:
• Workitem ITU-T focus group CarCom – acoustical
frontend for speech recognition
The Fully Networked Car
Geneva, 3-4 March 2010
15
ITU-T FG CarCOM: Hands-Free Telephony in Cars
16
Integrated systems:
• Completely integrated in the car infrastructure typically
including speech recognition, navigation….
After market systems:
• Independent of car infrastructure, sometimes including
speech recognition, navigation
Standardization activities in ITU based on work in FG
CarCOM:
•
•
•
ITU-T P.1100 for narrowband hands-free
ITU-T P.1110 for wideband hands-free
New work on subsystem requirements
The Fully Networked Car
Geneva, 3-4 March 2010
The Role of HEAD acoustics
o Providing expertise for testing and
optimization of all speech technologies used
in cars
o Providing test systems for speech
applications to the car industry, suppliers,
algorithm developers and chipset
manufacturers
o Supporting standardization since 20 years
based on the expertise and basic research at
HEAD acoustics
The Fully Networked Car
Geneva, 3-4 March 2010
17
Conclusion
18
o Speech technologies in cars may actively contribute to
o
o
o
o
deploy new services in cars
Speech technologies may help to reduce drivers
distraction if properly implemented
HEAD acoustics is providing all types of test services
and systems for testing and optimization of speech
technologies
ITU-T is an excellent source and basis for speech
related technologies and their standardization
FG CarCOM is actively working on advanced standards
for hands-free implementations and subsystems,
more:
http://www.itu.int/ITU-T/focusgroups/carcom/
The Fully Networked Car
Geneva, 3-4 March 2010