A path towards common quality assessment of narrowband and wideband voice

1
A path towards common quality assessment of
narrowband and wideband voice

• Trond Ulseth, Telenor RD
• (trond.ulseth_at_telenor.com)
• Workshop on Wideband Speech Quality in Terminals
  and Networks Assessment and Prediction
• 8th and 9th June 2004 - Mainz, Germany

2
Outline

• The history of wideband,
• Why has wideband not been a success (on ISDN)?
• A new era for wideband,
• What we (the telcos) can offer and what our
  customers need,
• The answer,
• A proposal for development of an extended
  (all-purpose) E-model.

3
The history of wideband

• 1988 CCITT published recommendation G.722,
• 1989/90 EuroISDN MoU signed,
• 1990 ISDN Test services in Europe,
• Tandberg Telecom launched an ISDN videophone at
  Telecom91 in Geneva. The videophone supported
  both narrowband and wideband audio,
• 1994 Videophones (and wideband telephony) was
  used to e.g. when interviewing athletes staying
  at the Olympic village at the Olympic Winter
  Games, Lillehammer, Norway,
• 1994 ISDN 7 kHz telephony teleservice defined by
  ETSI,
• 1999 ITU-T Recommendation G.722.1 published,
• 2002 First version of ITU-T Recommendation
  G.722.2 (AMR-WB) published.

4
The (missing) success of wideband telephony on
the ISDN

• Wideband telephony on the ISDN (the telephony 7
  kHz teleservice) has not been a success,
• Wideband voice for videotelephony/videoconferencin
  g has been a success,
• Narrowband telephony vs. Wideband telephony on
  the ISDN Is it the chicken and egg problem?

5
Telecom Innovators NewsMay 2004
HiFi VoIP Better than PSTN Voice
                         
By R. Brough Turner, SVP CTO, NMS VoIP
has been available in various forms since
199596, but significant adoption is a fairly
recent phenomenon. The most often cited reason
for this long incubation period is voice quality.
Even today, IT directors worry that IP-PBX voice
quality may not match expectations. This is
unfortunate as VoIP has the potential to offer
much better voice quality than the PSTN! Luckily,
examples of HiFi VoIP are appearing in the market
with additional products in the pipeline. Over
the next two years well see increasing
recognition of how HiFi VoIP overcomes the
quality problems of PSTN telephony.
6
VoIP/MMoIP a new era for wideband telephony?

• Most users dont care about the technology, they
  care about the quality/performance,
• VoIP has some similarities with the voice
  application of the videotelephoney and
  videoconferencing services there is not a
  separate service but the users can choose the
  coding algorithm (and quality),
• Global IP Sound has a wideband codec that is
  robust against packet loss. Products where this
  codec is implemented are available,
• Microsoft has included the Siren codec in Windows
  XP/ Windows Messenger.(Siren is the 16 kbit/s
  mode of the G.722.1 codec).

7
VoIP/MMoIP a new era for wideband
telephony?(more)

• The ITU-T Recommendation G.772.2 codec is
  standardised by 3GPP as the AMR-WB codec,
• Press reviews have emphasised the improved
  quality of some products (e.g. Skype),
• ConclusionWideband telephony will probably be a
  success, particularly for non-handset
  applications (headset and hands-free)

8
IP Networks Diverse quality

• Traditional IP networks are Best effort networks,
• To meet the requirements of real-time
  communication there has been a lot of work on
  implementing QoS mechanisms in IP networks,
• IP Network QoS classes are defined in ITU-T
  Recommendation Y.1541,
• A network QoS class definition is not sufficient
  for the user,
• Networks not belonging to the operator domain
  such as WLAN are not included,
• Media processing (voice coding) and other
  terminal related characteristics are not
  included.
• ETSI User Group recommends that there should be a
  standardized way to determine and present quality
  of service, system reliability and durability.
  This should include the development of
  standardized performance indicators.

9
Real-time service classes in an IP environment

• The network performance depends on factors such
  as network architecture, technology used and
  traffic,
• ITU-T Recommendation Y.1541 defines two real-time
  communication service QoS classes where the
  degree of interaction defines the delay
  requirement,
• ETSI TS 101 329-2 defines three speech QoS
  classes. However, the middle class (Narrowband)
  is divided into three sub-classes.The highest
  and lowest classes are
• Wideband (highest),
• Best effort (lowest).
• 3GPP TS 23.107 defines four QoS classes, one is
  the conversational class.

10
Our customers need a simple user perception based
classification

• The majority of our customers will not be able to
  assess the added effects of a number of
  parameters such as,
• Delay,
• Jitter,
• Packet loss,
• Dynamic range,
• Media bandwidth,
• Echo suppression,
• To define a single QoS class for e.g. voice tells
  the user that there is some mechanism in the
  network, but gives no information about the
  performance or a possibility to choose the
  desired performance.
• Conclusion We need an Index that is easy to
  understand. Telenor has been working on a Product
  Performance Index to describe the user perceived
  quality for different transmission technologies
  with a focus on data communication.A similar
  approach should be used for voice.

11
How to compare narrowband voice and wideband
voice?

• Subjective testing is of course the basis,
• However, there is a need for a tool based on
  objective measurements,
• The ultimate goal should be a calculation model
  where the performance of both narrowband voice
  and wideband voice could be estimated,
• A possible approach that has already been used by
  service providers and standardisation bodies is
  the E-model (ITU-T Recommendation G.107),
• The E-model has to extended to wideband to meet
  the requirements described.

12
The answer

• Is an extended1 E-model!

1 Applicable to both narrowband and wideband voice
13
The ITU-T E-model An important tool for
definition of CoS

• The initial work was carried out as a joint
  activity between the ETSI STCs BTC2, TE4 and TM5
  with TM5 being the lead STC. ETR 250 was
  published in 1996.
• The model has been adopted and refined by ITU-T
  SG 12.
• The aim is to give realistic, practical guidance
  rather than a scientifically exact treatment of
  quality factors. (Scope of ETR 250)
• The present version of the E-model is not
  addressing voice over IP, however, it is used in
  VoIP standards by both TIA and ETSI.

14
Extension of the E-model to cover wideband
telephony

• In the introduction of ETR 250 it is stated It
  (the ETR) may later be extended to cover
  non-handset telephony and wideband (7 kHz)
  telephony.
• ITU-T SG 12 is studying extensions of the
  E-model,
• One of the items of ITU-T Q8/12 is
  Investigation of the effects of wide-band
  transmission on speech sound quality and speech
  communication quality.In 4 years there has been
  no progress!
• Important questions are
• Is the approach the right one?
• What can be done to speed up the process?

start
15
Available wideband standards

• ITU-T Recommendations P.311 and P.341 specifies
  the transmission characteristics of wideband
  handset and hands-free terminals,
• Annex G to ITU-T Recommendation P.79 defines the
  wideband loudness rating calculation algorithm,
• butThere is a significant gap in the wideband
  voice standardisation compared with the
  narrowband voice standardisation.

16
Discussion of the elements of the E-model

• The E-modelR Ro Is Id Ie A
• Due to the extended low frequency bandwidth it is
  likely that Ro will be influenced by the room
  noise,
• Among other factors Is is influenced by the
  actual codec,
• The effect of delay on speech interactivity is
  probably not different for wideband speech
  compared with narrowband speech. However, the
  echo effects might be a challenge, particularly
  because the echo characteristics probably are
  different for a wideband terminal.
• The equipment impairment factor Ie is codec
  related

17
The path

• Wideband and narrowband should be considered as a
  single case, wideband should not be a special
  case.
• It is recommended to approach to problems on a
  step-by-step basis,
• To make a tool available as soon as possible, an
  intermediate phase is implemented,
• An analysis phase identifies elements of the
  E-model that are not affected by bandwidth and
  sets up a priority list for those elements that
  have to be modified,
• An implementation phase where the extended
  (all-purpose) E-model is developed.

18
The steps

• The equipment impairment factor of relevant
  wideband codecs should be estimated using the
  principles of ITU-T Recommendations P.833 or
  P.834,
• A wideband advantage factor is introduced as an
  intermediate tool,
• The intermediate all purpose (narrowband and
  wideband) E-model R Ro Is Id Ie A
  WBa
• The elements of the E-model are analysed. Those
  that need modifications to include wideband are
  identified,
• A priority list is set up,
• The required extensions to the E-model are
  implemented, following the priority list.

19
The intermediate wideband E-model

• The performance is estimated using,
• The existing E-model,
• The equipment impairment factor of the wideband
  codec used,
• The wideband advantage factor,
• This model could be described using one of the
  following alternatives,
• An Annex to ITU-T Recommendation G.107,
• A supplement to the ITU-T G-series
  Recommendations,
• An ETSI Technical Report.

20
Sometimes our best is simply not enough We have
to do what is required. Sir Winston Churchill
Thank you!