Presents
2006 IMTC Forum
ITU-T Workshop
G.729EV: An 8-32 kbit/s scalable wideband
speech and audio coder bitstream
interoperable with G.729
Presented by Christophe Beaugeant
On behalf of ETRI, France Telecom,
Matsushita, Mindspeed, Siemens
G.729EV Standardization history
•
•
•
Standardization work conducted in SG 16, WP 3, Question 10
(Rapporteur: Claude Lamblin)
T0 (approval of Term of References) declared in November 2004
Qualification test results in July 2005
– ETRI, France Telecom, Matsushita, Mindspeed, Siemens and VoiceAge
qualified
– These companies collaborated together to provide a single candidate
•
•
•
Optimization phase 08/05-01/06
Quality assessment tests (02-03/06): G.729EV passed all the
requirements.
Current status: under AAP procedure at the ITU-T
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
High level overview
•
8-32 kbit/s scalable wideband (50-7000 Hz) extension of G.729
– Wideband: More natural voice, better intelligibility, enhance
communication comfort
– Scalability : flexibility; dynamic, simple and low cost bit rate
adaptation
– Robustness : packet loss protection and concealment integrated
•
Build upon a three-stage structure, with 12 embedded layers:
– layers 1-2 : CELP, 50-4000 Hz, 8 -12 kbit/s
– Layer 3 : TDBWE, 50-7000 Hz, 14 kbit/s
– Layers 4-12: TDAC, 50-7000 Hz, 14 to 32 kbit/s
8
G.729a+
12
EXT.
32
14
TDBWE
TDAC
TDAC
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
G.729EV bitrate flexibility
8
12
G.729a+
EXT.
Original spectrum
Encoded spectrum
2
1
Narrow band
•
•
3
4
5
6
7
Wide band
Embedded CELP stage generates layers 1 and 2 : narrowband
synthesis (50-4000 Hz) at 8 and 12 kbit/s.
Layer 1 compliant with G.729 bitstream
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
G.729EV bitrate flexibility
12
G.729a+
14
TDBWE
EXT.
Original spectrum
Encoded spectrum
2
1
Narrow band
•
•
3
4
5
6
7
Wide band
Layer 3: Wideband Extension (TDBWE)
Wideband output (50-7000 Hz) at 14 kbit/s
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
G.729EV bitrate flexibility
12
G.729a+
14
24
TDAC
TDBWE
EXT.
32
TDAC
Original spectrum
Encoded spectrum
2
1
Narrow band
•
•
•
•
3
4
5
6
7
Wide band
Layers 4 to 12: Predictive transform coding referred as TDAC
Quality improvement from 14 to 32 kbit/s.
Encoding of the weighted CELP coding error in 50-4000 Hz band
Encoding input signal in 4000-7000 Hz band.
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Encoder:
High level overview
FEC encoder
S LB (n)
H1(z)
2
Hh1(z)
+
+ -
embedded
CELP encoder
(8-12 kbit/s)
S enh (n)
Aˆ ( z )
M
WLB(z)
S WB (n)
U
X
MDCT
TDAC
encoder
(16-32 kbit/s)
MDCT
H2(z)
2
(-1)n
Hh2(z)
S HB (n)
TDBWE
encoder
(14 kbit/s)
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Decoder:
High level overview
HPF
embedded
CELP decoder
(8-12 kbit/s)
+ +
+
adaptive
postfiltering
echo
Sˆ HB
( n)
2
G1(z)
echo
dˆ LB
( n)
pre/post-echo
reduction
D
WLB(z)-1
E
M
U
TDAC
decoder
(14-32 kbit/s)
Sˆ WB (n)
+
+
MDCT-1
+
MDCT-1
Sˆ HB (n)
X
MDCT
TDBWE
decoder
(14 kbit/s)
pre/post-echo
reduction
Sˆ
echo
LB
(-1)n
( n)
2
G2(z)
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Quality of G.729EV
•
Listening tests (02-03/06): G.729EV passed all the requirements.
•
Advanced embedment of three technologies:
– CELP : Good performance for Speech
– Bandwidth Extension: reduced bit rate to switch to wideband
– Transform Coding: high wideband quality with music and non
speech signals
•
Quality equivalent or better than other speech coding standard
– Optimum "state of the art" wideband quality for all types of signals
(including music) at 32 kbps
– High wideband quality of speech maintained at 14 kbps
– Narrow band "PSTN quality" at 12 kbps (equivalent to G711)
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Subjective tests
Clean speech in Narrowband in French
5
4,5
4
3,5
3
2,5
Direct
G.729A
CuT-8 k
G.729E
CuT-12 k
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Subjective tests
Wideband clean speech in French
5
4,5
4
3,5
3
Direct
G.722.2–
8,85k
G.722 –
48k
CuT-14 k
G.722 –
56k
CuT-24 k
CuT-32 k
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Objective Measurement
PESQ SCORES FOR CUT (MOS-LQO)
4.10
4.00
3.90
3.80
All Languages
English
French
German
Japanese
Korean
3.70
3.60
3.50
3.40
3.30
3.20
14k
16k
18k
20k
22k
24k
26k
28k
30k
32k
Bitrates
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Subjective tests
Wideband music at 32 kbit/s
5
4,5
4
3,5
3
Direct
G.722 @ 56 k
CuT-32 k
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Complexity of G.729EV
•
Complexity measured using ITU-T Software Tool Library
STL2005 v2.1 .
•
Complexity figures of the G.729EV coder (encoder/decoder).
Computational complexity
Static RAM
35.79 WMOPS
5 kwords
Scratch RAM
3.7 kwords
Data ROM
8.5 kwords
Program ROM
32 kwords
Delay
48.9375 ms
Low delay mode (narrow-band)
25 ms
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Foreseen application
•
Designed for packetized wideband voice (VoIP, VoATM, IP phone,
private networks)
– Applications requiring scalable wideband on top of G.729,
– Easy integration with existing VoIP infrastructure and services, fast
deployment
•
Extension to Multimedia services (Videoconference, VOD…)
– High quality with graceful degradation from WB (face-to-face) quality to
NB (telephone) quality
•
Advantage of Scalability:
–
–
–
–
–
Simple and low cost bit rate adaptation (very limited CPU)
Design for devices (gateways) that multiplex data/audio streams
Handling heterogeneous accesses/terminals
Optimization of bitrate to network capabilities (Network congestion)
Easy modification of bitrate by intelligent router (adjustment to network
capabilities)
– Voice messaging: capacity vs quality trade-off, access adaptation
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
WB VoIP & Conversational
e-Learning service
•
•
•
Excellent voice quality to fixed users
Fittest quality and service connectivity to mobile users
Trial service will be deployed by KT in Q3/2006
Service & QoS
Manager
Conversational
e-Learning
Contents
G.729EV
32kbits/s
IP
Backbone
Wireline(xDSL, FTTH, Cable)
Fixed User
G.729EV
32kbits/s
Public Wireless LAN
WiBro (Portable Internet)
Mobile Users
G.729EV
8~32kbits/s
For fittest
quality
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Enterprise Wideband VoIP
•
•
•
Excellent voice quality among G.729EV terminals
Seamless interoperability of legacy G.729 VoIP users and facilities
Anticipated companies: Dacom, Netcodec, Essetel,Onnet,,
Enterprise – Headquarter (Seoul)
IEEE802.11 b/g/e
G.729EV WiFi Phone or
G.729EV WiFI + Mobile
G.729EV IP phone
IEEE802.11 b/g/e
AAA, Call Agent
QoS Manager
AP
IEEE
802.1
Enterprise – Branch A (LA)
AP
Flow-based
QoS IP
Network
p/q
G.729 WiFi Phone
80 2. 1
E
E
E
I
p/q
G.729EV Soft-phone
Existing VoIP Facilities
G.729-G.711 Gateways
T1/E1
PSTN
Analog phones (Seoul)
T1/E1
PSTN
Analog phones (LA)
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Intelligent service
•
•
•
Conference call with heterogeneous terminals
Adaptability to terminal capabilities
Adaptability to Up-link / Downlink voice Quality requirement
Service
Manager
AP
G.729 EV Wifi phone
G729 EV Wifi + Mobile
Fixed User
24 kbit/s
14 kbit/s
14 kbit/s
32 kbit/s
14 kbit/s
12 kbit/s
12 kbit/s
32 kbit/s
G.729 EV
IP phone
G.729-G.711 Gateways
PSTN
Analog
Phone
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Conclusion
• G729 EV : 8-32 kbit/s scalable codec for narrowband
speech (8-12 kbit/s), wideband speech (12-32 kbit/s)
and audio (32 kbit/s)
• Good quality compared to existing codecs
• Applications foreseen used the advantage of scalability
• Codecs obtained through collaboration of ETRI,
France Telecom, Matsushita, Mindspeed, Siemens and
VoiceAge.
• Quick development to fit to operators’ needs in the
field applications over IP
• Extension: Stereo coding, Super Wideband (32 kHz)
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Back-up
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Subjective tests
Interfering Talker Background noise at 15 dB in NB in German
5
4,5
4
3,5
3
Direct
G.729A
CuT-8 k
CuT-12 k
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA
Subjective tests
Interfering talker background noise at 15 dB in WB in English
5
4,5
4
3,5
3
G.722 – 48k
G.722 – 56k
CuT-24 k
CuT-32 k
Direct
Joint IMTC Forum and ITU-T Workshop – “H.323, SIP: is H.325 next?” – 9-11 May 2006 – San Diego, CA, USA