International Telecommunication Union
The VDA Specification for car hands-
free: A step towards improved speech
quality in car type environments.
Balduin, W. (Ford AG); Best, H.(Adam Opel AG); Dietz, M.
(BMW); Eisner, G. (Volkswagen AG); Gierlich, H.W. (HEAD
acoustics GmbH); Hanke, A. (Audi AG); Heise, B. (Porsche AG);
Schaible, M. (DaimlerChrysler AG)
ITU-T Workshop on Standardization in Telecommunication for motor vehicles
ITU Headquarters, 24-25 November 2003
Outline
ITU-T
o Speech Quality: The Relevant Parameters
o Signal Processing in Car Hands-Free Systems
o The VDA Approach:
• Test Setup
• Tests
• Test Signals
o Requirements for further Work and
Standardization
2
dates
Speech Quality from the Users
ITU-T Perspective
talking
situation
conversational
speech situation
quality
listening
situation
3
dates
Choice and Design of Subjective
Tests Methods
ITU-T factors
that cause difficulties
in talking (type B) factors
that cause difficulties
in conversation (type C)
factors
that cause difficulties
in listening (type A)
parameter identification
and definition correlation
test
to instrumental
design
laboratory tests
value selection
quality assessment
quality prediction
under realistic
network conditions
Comparison
of implementations data base
for further tests
4
dates
Parameters Relevant to Quality
ITU-T
o Auditory perceived parameters which
determine the speech quality of hands-free
terminals:
• Quality of background noise transmission
• (Speech) sound quality
• Delay and echo
• Double talk capability
• Switching and echo during double talk
• Loudness
• (System) noise
5
dates
Hands-Free Signal Processing
ITU-T
Noise-
reduction
EC Attenuation Coder
Microphon –
(Array) -
NLP
NR Coder
Modulation
H(f )
/
Demodulation
+ Decoder
Loudspeaker EC e.g. GSM - Modul
CN
Comfort Decoder
Noise
Speech Recognition
6
dates
VDA: The Test Setup
ITU-T
• real car cabin Loud-
speaker
• installed hands-free equalizer
Background
terminal noise
(HF)
System HFT
• System simulator Simulator
(GSM, ...)
Measure-
• Background noise ment
simulation system
• HATS Head and
Torso Simulator
(ITU-T P.58, P.581) Subwoofer
• VDA Specification or
HQS-HFT
7
dates
The Structure of the VDA
ITU-T Specification
o Microphone measurements:
• Sensitivity, frequency responses, ambient
noise reduction…
o Hands-free measurements:
• Delay
• Standard parameters (LR, responses, noise,
out of band …)
• Echo related parameters (TCL, vs. time,
spectral…)
• Switching characteristics
• Double talk parameters (attenuation, echo)
• Background noise transmission parameters
8
dates
Microphone Measurements
ITU-T Two test principles:
Measurements
in anechoic Measurements
test conditions in the car cabin
• sensitivity • Sensitivity
• linearity • linearity
• distortions • distortions
• frequency response
• idle noise
• ambient noise reduction 9
dates
Hands-free tests
ITU-T
Standard-
parameters Advanced tests in
in the car the car
! Delay
! Background Noise suppression
(DeLSM
! Frequency Response ! Distortions
! Loudness Ratings ! Echo loss - time variant,
! Distortions spectral
! Noise ! Double talk performance
! Out of Band signals ! Background noise performance
! Terminal Coupling Loss ! Comfort noise insertion
10
dates
Testsignals:
ITU-T
Speech-like, but adapted to the specific measuerement task:
„Composite Source“ Signals
Artificial Voice
Modulated
Voiced
Sound
11
dates
Testsignals (CSS)
ITU-T
Composite source Signals
Pseudo Noise
Voiced Sound
Pause
Signal description in
ITU-T P.501
12
dates
(Noise)- Measurement with
ITU-T Activation Test Signal
Test signal
activation CSS
-28.7 dBPa
Analysis
Test / Analysis
after activation
window length 1s
(smaller if
necessary!)
Requirements
GSM 40 dB (3GPP)
> 46 dB
recommended (ITU)
> 33 dB at maximum
volume
14
dates
Double Talk Performance
ITU-T
Test signal
Combination of
CS signals analysis single talk
Test / Analysis
Level vs. time
Requirements
Type of
classification!
15
dates
Classification based on ITU-T
ITU-T Rec. P.340
Type 1 Type 2 Type 3
Subjectively Full duplex Partial No duplex
Relevant capability duplex capability
Parameter capability
(in P.340 and
VDA more
precise)
Objective
Parameter
• aH,SND,DT 12 dB
• aH, RCV,DT 10 dB
• Echo ≥ 37 dB 36 - 21 dB < 21 dB
attenuation
16
dates
Summary
ITU-T
o Speech quality has to consider the
conversational situation - the listening
situation, the talking situation and double talk
o Objective parameters are combined by
standard parameters as well as specific
parameters determining double talk and
background noise transmission quality
o Specific test signals are suited for the detailed
evaluation of the various parameters
17
dates
To Do:
ITU-T Requirements for Future Work
o Speech quality in the presence of
background noise
o Quality of the noise transmission
o Dynamic background noise simulation
o Speech quality in combination with
varying network quality
o Requirements for wideband
conversational speech quality
18
dates