ETSI TR 103 138 V1.4.1 (2016-09)

Speech and multimedia Transmission Quality (STQ); Speech samples and their use for QoS testing

RTR/STQ-00215m

General Information

• Status: Published
• Publication Date: 19-Sep-2016
• Technical Committee: STQ MOBILE - STQ Mobile
• Current Stage: 12 - Completion
• Due Date: 11-Oct-2016
• Completion Date: 20-Sep-2016

Standards Content (Sample)

ETSI TR 103 138 V1.4.1 (2016-09)






TECHNICAL REPORT
Speech and multimedia Transmission Quality (STQ);
Speech samples and their use for QoS testing

�

---------------------- Page: 1 ----------------------
2 ETSI TR 103 138 V1.4.1 (2016-09)



Reference
RTR/STQ-00215m
Keywords
QoS, quality, speech
ETSI
650 Route des Lucioles
F-06921 Sophia Antipolis Cedex - FRANCE

Tel.: [phone]  Fax: [phone]

Siret N° 348 623 562 00017 - NAF 742 C
Association à but non lucratif enregistrée à la
Sous-Préfecture de Grasse (06) N° 7803/88

Important notice
The present document can be downloaded from:
http://www.etsi.org/standards-search
The present document may be made available in electronic versions and/or in print. The content of any electronic and/or
print versions of the present document shall not be modified without the prior written authorization of ETSI. In case of any
existing or perceived difference in contents between such versions and/or in print, the only prevailing document is the
print of the Portable Document Format (PDF) version kept on a specific network drive within ETSI Secretariat.
Users of the present document should be aware that the document may be subject to revision or change of status.
Information on the current status of this and other ETSI documents is available at
https://portal.etsi.org/TB/ETSIDeliverableStatus.aspx
If you find errors in the present document, please send your comment to one of the following services:
https://portal.etsi.org/People/CommiteeSupportStaff.aspx
Copyright Notification
No part may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying
and microfilm except as authorized by written permission of ETSI.
The content of the PDF version shall not be modified without the written authorization of ETSI.
The copyright and the foregoing restriction extend to reproduction in all media.

© European Telecommunications Standards Institute 2016.
All rights reserved.

TM TM TM
DECT , PLUGTESTS , UMTS and the ETSI logo are Trade Marks of ETSI registered for the benefit of its Members.
TM
3GPP and LTE™ are Trade Marks of ETSI registered for the benefit of its Members and
of the 3GPP Organizational Partners.
GSM® and the GSM logo are Trade Marks registered and owned by the GSM Association.
ETSI

---------------------- Page: 2 ----------------------
3 ETSI TR 103 138 V1.4.1 (2016-09)
Contents
Intellectual Property Rights . 4
Foreword . 4
Modal verbs terminology . 4
Introduction . 4
1 Scope . 5
2 References . 5
2.1 Normative references . 5
2.2 Informative references . 5
3 Abbreviations . 6
4 Devices and network access . 6
4.1 Mobile devices . 6
4.2 ISDN/PSTN . 7
4.3 Test scenarios . 7
4.3.1 General aspects . 7
4.3.2 Narrowband telephony and narrowband test scenario . 7
4.3.3 Wideband telephony and super-wideband test scenario . 8
5 Speech samples . 9
5.1 General aspects . 9
5.2 Pre-filtering of speech signals . 9
5.2.1 Emulation of handsets . 9
5.2.2 Filter for narrow-band test scenarios . 9
5.2.2.1 IRS send Filter . 9
5.2.2.2 MSIN Filter . 10
5.2.2.3 Recommended filters to use in narrowband mobile test scenarios . 11
5.2.3 Filter for wideband and super-wideband telephony test scenarios. 11
5.2.3.1 14 kHz bandpass . 11
5.2.3.2 Recommendation ITU-T P.341 . 11
5.2.3.3 Recommended filters to use in super-wideband mobile test scenarios . 11
5.2.4 Reference signals . 12
5.3 Audio level . 12
5.3.1 Nominal level . 12
5.3.2 Level adjustment with Recommendation ITU-T P.56 . 12
5.3.3 Input level at test devices . 12
6 Scenarios . 13
6.1 Narrowband-Measurement Land to Mobile . 13
6.2 Narrowband-Measurement Mobile to Land . 13
6.3 Mobile to Mobile . 14
6.3.1 Narrowband . 14
6.3.2 Wideband and super-wideband . 14
7 Synopsis . 15
Annex A: Coefficients for the reconstruction lowpass filter . 16
Annex B: Bibliography . 17
Annex C: Speech Samples . 18
C.1 Introduction . 18
C.2 Design. 18
C.3 Example results . 18
C.4 Technical specification . 20
History . 22

ETSI

---------------------- Page: 3 ----------------------
4 ETSI TR 103 138 V1.4.1 (2016-09)
Intellectual Property Rights
IPRs essential or potentially essential to the present document may have been declared to ETSI. The information
pertaining to these essential IPRs, if any, is publicly available for ETSI members and non-members, and can be found
in ETSI SR 000 314: "Intellectual Property Rights (IPRs); Essential, or potentially Essential, IPRs notified to ETSI in
respect of ETSI standards", which is available from the ETSI Secretariat. Latest updates are available on the ETSI Web
server (https://ipr.etsi.org/).
Pursuant to the ETSI IPR Policy, no investigation, including IPR searches, has been carried out by ETSI. No guarantee
can be given as to the existence of other IPRs not referenced in ETSI SR 000 314 (or the updates on the ETSI Web
server) which are, or may be, or may become, essential to the present document.
Foreword
This Technical Report (TR) has been produced by ETSI Technical Committee Speech and multimedia Transmission
Quality (STQ).
Modal verbs terminology
In the present document "should", "should not", "may", "need not", "will", "will not", "can" and "cannot" are to be
interpreted as described in clause 3.2 of the ETSI Drafting Rules (Verbal forms for the expression of provisions).
"must" and "must not" are NOT allowed in ETSI deliverables except when used in direct citation.
Introduction
Conducting drive test in multi technology environment presents a challenge to all parties. And the complexity and
variance of the different scenarios need to be broken down to handy instructions for those who actually configure and
conduct the measurements, such as Network Operators, Service Providers, Equipment Vendors and Regulatory
Authorities.
ETSI

---------------------- Page: 4 ----------------------
5 ETSI TR 103 138 V1.4.1 (2016-09)
1 Scope
The present document introduces and explains the use and application of speech samples to determine the objective
listening quality (LQO) in narrowband (NB), wideband (WB) and super-wideband (SWB) for different scenarios such
as connections between fixed networks and mobile terminals.
2 References
2.1 Normative references
Normative references are not applicable in the present document.
2.2 Informative references
References are either specific (identified by date of publication and/or edition number or version number) or
non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the
reference document (including any amendments) applies.
NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee
their long term validity.
The following referenced documents are not necessary for the application of the present document but they assist the
user with regard to a particular subject area.
[i.1] Recommendation ITU-T P.48: "Specification for an intermediate reference system".
[i.2] Recommendation ITU-T P.800: "Methods for subjective determination of transmission quality".
[i.3] Recommendation ITU-T P.830: "Subjective performance assessment of telephone-band and
wideband digital codecs".
[i.4] Recommendation ITU-T P.862: "Perceptual evaluation of speech quality (PESQ): An objective
method for end-to-end speech quality assessment of narrow-band telephone networks and speech
codecs".
[i.5] Recommendation ITU-T P.862.1: "Mapping function for transforming P.862 raw result scores to
MOS-LQO".
[i.6] Recommendation ITU-T P.862.2: "Wideband extension to Recommendation P.862 for the
assessment of wideband telephone networks and speech codecs".
[i.7] Recommendation ITU-T P.862.3: "Application guide for objective quality measurement based on
Recommendations P.862, P.862.1 and P.862.2".
[i.8] Recommendation ITU-T P.863: "Perceptual objective listening quality assessment (POLQA)".
[i.9] Recommendation ITU-T P.863.1: "Application Guide for the Recommendation ITU-T P.863".
[i.10] Recommendation ITU-T G.711: "Pulse code modulation (PCM) of voice frequencies".
[i.11] Recommendation ITU-T G.191: "Software tools for speech and audio coding standardization".
[i.12] Recommendation ITU-T P.341: "Transmission characteristics for wideband digital loudspeaking
and hands-free telephony terminals".
[i.13] Recommendation ITU-T P.56: "Objective measurement of active speech level".
[i.14] Recommendation ITU-T P.501: "Test signals for use in telephonometry".
ETSI

---------------------- Page: 5 ----------------------
6 ETSI TR 103 138 V1.4.1 (2016-09)
3 Abbreviations
For the purposes of the present document, the following abbreviations apply:
AMR Adaptive Multi-Rate codec
AMR-WB Adaptive Multi-Rate codec Wide Band
ASL Active Speech Level
EFR Enhance Full Rate codec
EVS Enhanced Voice Services, speech codec
FIR Finite Impulse Response filter
IRS Intermediate Reference System
ISDN Integrated Services Digital Network
LQO Listening Quality Objective
MOS Mean Opinion Score
MSIN Mobile Station Input filter
NB Narrow Band
NTP Network Terminating Point
OVL Overload point
PBX Private Branch Exchange
PC Personal Computer
PCM Pulse Code Modulation
PSTN Public Switch Telecommunication Network
SWB Super Wide Band
VoLTE Voice over LTE
WB Wide Band
4 Devices and network access
4.1 Mobile devices
There are only a few devices and access interfaces that play a role in end-to-end mobile network testing. In end-to-end
testing a test connection between two endpoints is established. This determines the access interfaces and devices.
The mobile device is not a pure access device to the mobile network. It contains complex components for speech
processing and becomes therefore an important contributor to the overall quality measured in the test connection.
Mobile devices do not have a standardized audio interface, neither digital nor analogue. As common practice the
headset connector of the mobile device is used as access interface for audio insertion and capturing. As a pre-condition
for audio insertion and capturing, the measurement equipment has to match to the devices headset connector in
impedance and level.
It has to be noted that in this setup the mobile devices are used in headset mode. Devices apply individual audio
profiles, means individual settings in filtering, amplification and noise- and echo treatment for connected headphones or
the use of the internal microphone. Often there is a third mode that applies when a handsfree loudspeaker set is
connected. Since the audio processing in headphone mode is different from the use of internal microphone, such a test
connection emulates a user with a headphone (personal handsfree kit) connected by wire to the headphone connector.
ETSI

---------------------- Page: 6 ----------------------
7 ETSI TR 103 138 V1.4.1 (2016-09)
4.2 ISDN/PSTN
ISDN or (analogue) PSTN interfaces are not directly belonging to the mobile network but they are usually used as
defined endpoint of the test connection. As access point to the ISDN or PSTN network a real consumer telephone
device is not used but rather an ISDN or PSTN interface module as e.g. a PC card. It enables an electrical connection to
the network for audio transmission and processes all the signalling information. The interface module or PC card is
usually accessed with a digitalized speech signal in PCM format. The format is preferably 16 bit or 13 bit linear PCM
sampled at 8 kHz or 16 kHz. Some interfaces expect 8 bit A-Law PCM that can be used in case of ISDN but is not
recommended for PSTN, since it will cause an additional A-Law compression step in the test connection.
NOTE: The A-Law signal would be decompressed and fed as analogue signal in the local loop, where the regular
A-Law compression will be at the digital NTP or the PBX.
Today, ISDN/PSTN channels are narrow-band only. Thus, a transmission to an ISDN/PSTN end-point is always
restricted to narrowband despite that the airlink can use AMR-WB. The transition to narrowband is part of the gateway
to the ISDN/PSTN.
4.3 Test scenarios
4.3.1 General aspects
The analogue circuits of almost all mobile devices are able to process wideband or fullband speech. Whether a call is
transmitting narrowband or wideband or above speech depends on the wideband coding capability of the phone, the
network and call setup. The subscriber cannot control whether the phone connects in narrowband, in wideband or in
super wideband. The established channel determines the transmission bandwidth of the channel that can be narrowband,
wideband, super-wideband or even fullband.
4.3.2 Narrowband telephony and narrowband test scenario
The conventional narrowband or normal-band telephony is traditionally using a pass-band from 300 Hz to 3 400 Hz. In
digital transmission the technical limit is given by the Nyquist frequency due to sampling at 4 kHz upper audio
transmission limit; there is no limit at the lower boundary. Today's narrowband speech codecs as EFR or AMR are also
able to encode an audio band up to 4 kHz. Despite that fact, in practice a dedicated filtering is applied to the signal.
Usually, there is a bandpass that is wider than the traditional pass-band but still limiting at the lower and upper range.
The actual transmission characteristic is depending on the phone manufacturer and the setting of the phone. There are
no binding limits or characteristics.
Testing narrowband is not tied to a narrowband channel. Narrowband testing means that the listening quality is
estimated as listening through a conventional handset, the objective quality model filters the signal with such a
band-pass and compares the speech signal to an ideal narrowband reference signal too. This restriction to a narrowband
bandpass is applied despite the fact of the signal bandwidth passed through the channel.
For testing a narrowband scenario using a mobile access device there are two setups:
1) Insertion of a signal that exceeds the traditional narrowband bandwidth, e.g. 50 Hz to 3 800 Hz or even 50 Hz
to 8 000 or 50 Hz to 14 000 Hz. In this case, the limitation of the signal is done by the device and the channel,
while the device usually limits at most. At the receiving side, the recorded speech signal is compared to an
ideal narrowband signal (at a bandwidth of 50 Hz to 3 800 Hz). In this test case the filter characteristic of the
mobile device used has a significant influence on the estimated quality, since all restrictions to the reference
bandwidth are considered as degradation. The predicted MOS describes the overall quality as it is perceived by
the particular device and the channel; the score is device dependent.
2) Insertion of a signal that emulates a traditional sending path that is close to the defined passband of 300 Hz to
3 400 Hz. Therefore the test speech signal is filtered with a bandpass filter as e.g. IRSsend or MSIN. Usually,
those filters are narrower than the phone's characteristic. The phone's band limitations will not affect
significantly the speech signal anymore. By using such a setup, the filter characteristic of the particular phone
becomes less influencing. The bandwidth of the signal at receiving side is than widely dominated by the
applied pre-filtering and widely the same for all devices. The estimated score becomes less phone dependent.
ETSI

---------------------- Page: 7 ----------------------
8 ETSI TR 103 138 V1.4.1 (2016-09)
The approach (1) is recommended for device testing. For field testing of mobile network quality the setup (2) is
recommended. It focuses more on network quality than on device depending audio filtering.
Please note that the term narrowband test scenario does not depend on the actual transmission capability of the channel
but rather on the quality reference that is just narrowband. Even a wideband channel can be tested in a narrowband
setup, it can be compared to listening wideband with a traditional handset, the upper frequency ranges are just not
perceptible by such a transducer.
Typical MOS scores in a narrowband scenario are:
• 4.5 for a complete transparent narrowband signal.
• 4.4 for an ISDN signal (coded with Recommendation ITU-T G.711 [i.10] A-Law).
• 4.2 to 4.3 for a perfectly processed signal with AMR at 12,2 kbit/s.
• 3.4 to 3.6 for a perfectly processed signal with AMR at 4,75 kbit/s.
Quality testing in a narrowband test scenario is used for a long time and most of published MOS scores relate to this
scenario. The established Recommendation ITU-T P.862.1 [i.5] is an objective measure emulating a narrowband
scenario. Also, the new Recommendation ITU-T P.863 [i.8] supports a dedicated narrowband test modus, where signal
predictions are made according to a narrowband test setup.
4.3.3 Wideband telephony and super-wideband test scenario
For wideband telephony typically a transmission capability of 100 Hz to 7 000 Hz is defined. Similar to narrowband,
the technical limits for a wideband transmission and channel are from often 50 Hz to 8 000 Hz due to the sampling
frequency of 16 000 Hz.
NOTE: The AMR-WB speech codec limits at 6 400 Hz itself due to an internal sampling frequency of 12,8 kHz.
The step beyond wideband is called super-wideband and enables a transmission bandwidth from 50 Hz to 14 000 Hz. In
practice, super-wideband can be seen as full-band for human speech, since there are no relevant signal parts above
14 000 Hz.
The recently standardized EVS speech codec supports all audio bandwidths from narrow-band, wideband, super-
wideband and even to full-band. In comparison with AMR and AMR-WB, which can adapt bitrate but support only one
fix audio bandwidths. The EVS speech codec can change both, audio bandwidth and bitrate and is able to choose to the
best compromise between bitrate and audio bandwidth adaptively. For VoLTE the EVS codec will support super-
wideband audio as default.
A wideband or super-wideband transmission requires a corresponding channel and two endpoint devices, those are able
to process wideband or super-wideband speech. Today, wideband in the field can only be tested in mobile to mobile
connections, since ISDN/PSTN are restricted to narrowband.
In a traditional wideband scenario, a wideband signal becomes compared to an ideal 100 Hz to 7 000 Hz or 50 Hz to
8 000 Hz signal. However, the there is a tendency to evaluate and score traditional wideband directly by comparing to
super-wideband signal as an ideal reference. Along with the standardization of Recommendation ITU-T P.863 [i.8]
there is the super-wideband mode recommended, where the recorded signal is compared with a 50 Hz to 14 000 Hz
reference signal (Recommendation ITU-T P.862.2 [i.6] wideband supports a dedicated wideband modus, however this
measure was not established in the field and superseded by Recommendation ITU-T P.863 [i.8] super-wideband mode).
The super-wideband scenario can be imagined as listening through a high quality headphone without perceptible
restrictions in transmission. It is as a mono listening situation, where the same signal is perceived on both ears.
The actual limitation to 7 000 Hz or 8 000 Hz in a real wideband transmission as with the AMR-WB will lead to slight
degradation compared to a reference of 50 Hz to 14 000 Hz. For testing a super-wideband channel, the wideband and
super-wideband scenario is the best suited test scenario. In that scenario the signal can be evaluated completely up to its
upper spectral range of 14 000 Hz. Super-wideband mode gives the possibility to relate each limitation to an ideal
sample (quasi fullband reference) and can still be used in the future, when super-wideband codecs become deployed in
mobile networks.
ETSI

---------------------- Page: 8 ----------------------
9 ETSI TR 103 138 V1.4.1 (2016-09)
From a testing point of view, flat filtered super-wideband signal is inserted in the access interface. All limitations in
bandwidth applied to the signal are taken into account. Typical MOS scores in a super-wideband scenario are:
• 4.75 for a full transparent signal from 50 Hz to 14 000 Hz or more.
• 4.2 to 4.5 for a full transparent wideband signal from 50 Hz to 7 000 or 8 000 Hz.
• 3.8 to 4.1 for a transparent processing with AMR-WB 12,65 and no further limitations in bandwidth.
• 3.2 to 3.5 for a transparent processing with AMR 12,2 in narrowband.
5 Speech samples
5.1 General aspects
Starting from the original speech sample recorded in the studio the sample need to be processed before they can be used
in instrumental speech testing.
Speech samples for quality testing are usually composed by a subsequent series of sentences spoken by a human
speaker. Traditionally, a sentence pair of two sentences is used in auditory tests following Recommendation ITU-T
P.800 [i.2] and for instrumental testing as well.
Recommendations on recording and processing of speech samples for testing speech quality are given in
Recommendations ITU-T P.800 [i.2] and P.830 [i.3]. Speech samples to be used for instrumental testing of speech
quality have to fulfil additional technical requirements regarding temporal structure, noise floor and similar. Those
recommendations are given in Recommendations ITU-T P.862.3 [i.7] and P.863.1 [i.9].
Typically, there is a systematic difference in scoring male or female voices, where male voices are scored by
instrumental measures like P.862 [i.4] and P.863 [i.8]. For the purpose of automated testing as in drive test tools, speech
samples combining sentences spoken by a male and a female talker is a preferable solution.
5.2 Pre-filtering of speech signals
5.2.1 Emulation of handsets
Depending on the application to be tested different filters need to be applied. In this context, filtering applies to an
upfront filtering applied to the speech signal before it becomes inserted in the test device or the network interface
respectively. This filter emulates the transmission characteristic of the microphone and its connection circuit, which is
not present in an electrical insertion. After filtering, the signal becomes closer to the signal that would naturally be
available at this point of insertion.
5.2.2 Filter for narrow-band test scenarios
5.2.2.1 IRS send Filter
The IRS filter (IRS stands for Intermediate Reference System) emulates a transmission characteristic of a traditional
narrowband handset. There is an IRS send filter for the microphone and sending characteristic and an IRS receive filter
for the characteristic of the receiving side including a (electro-dynamic) transducer.
The IRS send filter can be imagined as a bandfilter slightly wider than the normal passband but with a significant
pre-emphasis towards 2 700 Hz. The classical IRS filters are defined in Recommendation ITU-T P.48 [i.1].
There is a revised characteristic (Modified IRS send) defined in Recommendation ITU-T P.830 [i.3] that has slightly
weaker roll-off characteristics at the band limits. The difference at the upper boundary becomes much smaller, when a
downsampling filter to 8 000 Hz is applied to the IRS filtered signal that is usual for input signals in a narrow-band
channel (figure 1).
ETSI

---------------------- Page: 9 ----------------------
10 ETSI TR 103 138 V1.4.1 (2016-09)
10
B
0
d
-10
-20
-30
IRS send MOD
-40
IRS send
-50
0 2000 4000 6000 8000
Hz

Figure 1: Frequency responses for IRS send and Modified IRS send filters
'Modified IRS send' is the pre-filter that is used by ITU-T for testing and evaluating narrowband speech codecs. IRS
send and Modified IRS send filters are provided as examples in Recommendation ITU-T G.191 [i.11] that is a
collection of processing algorithms of ITU-T.
5.2.2.2 MSIN Filter
The MSIN filter is also emulating a sending device but has no pre-emphasis (it is almost fla
...