Professional Documents
Culture Documents
Voice Quality Enhancement
Voice Quality Enhancement
GBSS12.0
Feature Parameter Description
Issue 01
Date 2010-06-30
and other Huawei trademarks are trademarks of Huawei Technologies Co., Ltd.
All other trademarks and trade names mentioned in this document are the property of their respective
holders.
Notice
The purchased products, services and features are stipulated by the contract made between Huawei and
the customer. All or part of the products, services and features described in this document may not be
within the purchase scope or the usage scope. Unless otherwise specified in the contract, all statements,
information, and recommendations in this document are provided "AS IS" without warranties, guarantees or
representations of any kind, either express or implied.
The information in this document is subject to change without notice. Every effort has been made in the
preparation of this document to ensure accuracy of the contents, but all statements, information, and
recommendations in this document do not constitute the warranty of any kind, express or implied.
Website: http://www.huawei.com
Email: support@huawei.com
GSM BSS
Voice Quality Enhancement Contents
Contents
1 Introduction ................................................................................................................................1-1
1.1 Scope ............................................................................................................................................ 1-1
1.2 Intended Audience ........................................................................................................................ 1-1
1.3 Change History.............................................................................................................................. 1-1
2 Overview .....................................................................................................................................2-1
3 VQE1.0 .........................................................................................................................................3-1
3.1 Acoustic Echo Cancellation ........................................................................................................... 3-1
3.2 Automatic Level Control ................................................................................................................ 3-2
3.3 Automatic Noise Compensation .................................................................................................... 3-3
3.4 Automatic Noise Restraint ............................................................................................................. 3-4
4 VQE3.0 .........................................................................................................................................4-1
4.1 ACLP ............................................................................................................................................. 4-1
5 Parameters .................................................................................................................................5-1
6 Counters ......................................................................................................................................6-1
7 Glossary ......................................................................................................................................7-1
8 Reference Documents .............................................................................................................8-1
1 Introduction
1.1 Scope
This document describes the voice quality enhancement (VQE) feature.
Document Issues
The document issues are as follows:
01 (2010-06-30)
Draft (2010-03-30)
01 (2010-06-30)
This is the first release of GBSS12.0.
Compared with issue draft (2010-03-30) of GBSS12.0, issue 01 (2010-06-30) of GBSS12.0 incorporates
the changes described in the following table.
Draft (2010-03-30)
This is the draft release of GBSS12.0.
2 Overview
Voice Quality Enhancement (VQE) is a collective term for the five voice quality enhancing features:
Acoustic Echo Control (AEC), Automatic Level Control (ALC), Automatic Noise Compensation (ANC),
Automatic Noise Restraint (ANR), and Anti-clip (ACLP). The enabling of each feature is controlled by a
corresponding parameter. The five features improve the user experience from different aspects.
VQE applies to only the A over TDM network.
3 VQE1.0
VQE1.0 contains four algorithm modules: AEC, ALC, ANC, and ANR. Each algorithm module improves
the voice quality from a specific aspect.
A speech channel can be simplified as a module with four ports: source in (Sin), source out (Sout),
remote in (Rin), and remote out (Rout). Figure 3-1 shows the relations between four algorithm modules
of VQE1.0.
Figure 3-1 Relations between four algorithm modules of VQE1.0
CODEC VQE
Sin A A A Sout
DEC E N L
C R C
BTS-MS PSTN
Rout A A A Rin
ENC N L N
C C R
BSC-TC
The uplink data Sin after decoding is sent to the AEC module, ANR module, and then ALC module in
succession for acoustic echo control, automatic noise restraint, and automatic level control, and finally
Sout is obtained.
The downlink data Rin is sent to the ANR module, ALC module, and then ANC module in succession for
automatic noise restraint, automatic level control, and automatic noise compensation. Then after
encoding, the Rout is obtained.
The Acoustic Echo Cancellation (AEC) feature of the BSC is implemented by the DSP of the DPU board.
The DSP analyzes the uplink and downlink digital voice signals, searches for acoustic echoes in the
uplink speech signals, and suppresses the acoustic echoes. This feature simplifies the acoustic echo
positioning and improves the user experience. (GBFD-115602 Acoustic Echo Cancellation(AEC))
Principles
The AEC feature eliminates the uplink acoustic echoes based on the voice signal characteristics.
The AEC feature is enabled only when the AecEnFlag parameter is set to ON(On).
AEC performs the following operations to eliminate the uplink acoustic echoes:
1. AEC checks the downlink speech signals and stores the downlink speech signal characteristics.
2. AEC searches for the stored downlink speech signal characteristics based on the setting of the
AecTail parameter, and then compares the downlink speech signal characteristics with the uplink
speech signal characteristics to determine whether acoustic echoes exist in the uplink speech
signals.
− Ifacoustic echoes do not exist in the uplink speech signals or the acoustic echoes are much weaker
than the uplink speech signals, AEC does not process the uplink speech signals.
− If acoustic echoes exist in the uplink speech signals, AEC performs step 3.
3. The BSC further analyzes the uplink voice signals based on the parameter AecDefaultERL. If
downlink speech signals exist in the uplink speech signals, AEC gradually attenuates the acoustic
echoes. If the level of the downlink acoustic echoes in the uplink speech signals reaches a certain
value, AEC performs non-linear processing on the speech signals.
The echo generated in the call between an MS and a fixed-line phone is an electric echo, which is processed by the CN.
The echo generated in the call between two MSs is an acoustic echo, which is processed by the BSS.
If a special transmission mode is used in the network, for example, in a case where the satellite transmission is used
over the Abis interface or Ater interface, the round trip time (RTT) needs to be estimated and AecPureDelay needs to
be set. The maximum value for AecPureDelay is 1000 ms.
If the Tandem Free Operation (TFO) feature is enabled at both ends of a call, the AEC feature does not take effect for
this call.
If the voice volume in the network is stable, you are advised to adopt the default configuration of ALC.
Principles
There are three ALC modes: fixed gain mode, fixed level mode, and adaptive level mode.
The ALC feature is enabled only when the AlcEnFlag parameter is set to ON(On).
Fixed gain mode
When the AlcAdaptMode parameter is set to Fixed Gain Mode, the uplink and downlink voice data is
sent to the fixed gain module. ALC increases or decreases the amplitude of the voice level according
to the predefined AlcFixGain.
Fixed level mode
When the AlcAdaptMode parameter is set to Fixed Level Mode, the uplink and downlink voice data
is sent to the fixed level module. ALC increases or decreases the amplitude of the voice level
according to the predefined AlcFixLev to keep the voice level in compliance with the predefined
AlcFixLev.
Adaptive level mode
When the AlcAdaptMode parameter is set to Adaptive Mode, the uplink and downlink voice data is
sent to the adaptive level module. ALC increases or decreases the amplitude of the voice level
according to the predefined AlcMinLev and AlcMaxLev to keep the voice level within the range
defined by AlcMinLev and AlcMaxLev.
Generally, the ALC mode that best suits the actual configuration is adopted to process the uplink and
downlink voice data, thus achieving the gain control over the voice data flow.
ALC is effective for both uplink voice level and downlink voice level. Therefore, when the AlcAdaptMode parameter is
set to Fixed Gain Mode, the AlcFixGain parameter should not be set to a too great value to prevent too high to too low
voice level caused by the dual application of ALC.
The ALC Max Gain parameter specifies the maximum gain of the voice level. This parameter is valid only when the
AlcAdaptMode parameter is set to Fixed Level Mode or Adaptive Mode. By default, this parameter is set to 6 dB. If
the setting of the ALC Max Gain parameter affects the actualization of the target level in the fixed level mode or the
adaptive level mode, you are advised to set the ALC Max Gain parameter to the largest possible difference between
the voice level and the expected target level.
If the Tandem Free Operation (TFO) feature is enabled during a call, the ALC feature does not take effect for this call.
If the background noise is loud at the peer end of a call, the noise is transmitted to the local end in the downlink speech
signals, and ANC may increase the amplitude of the noise in the downlink speech signals. This adversely affects the voice
quality at the local end. In this case, you are advised to enable the ANC feature with the Automatic Noise Restraint (ANR)
feature at the same time. The ANR feature can reduce the noise in the uplink speech signals.
Principles
ANC processes only downlink speech signals. ANC adjusts the gain of downlink digital speech signals
according to the uplink input noise level every 20 ms to dynamically change the amplitude of digital
speech signals. This keeps the ratio of downlink output voice level to uplink input noise level in
compliance with a predefined value. Thus, the user at the local end can clearly hear the voice from the
other end over the mobile phone.
The ANC feature is enabled only when the AncEnFlag parameter is set to ON(On).
ANC performs the following operations to improve the voice quality:
1. Calculates the noise level in the uplink.
2. Calculates the voice level in the downlink.
− If
the ratio of the voice level in the downlink to the noise level in the uplink is greater than or equal to
AncSnrGateRS, ANC does not process the downlink speech signals.
− If
the ratio of the voice level in the downlink to the noise level in the uplink is lower than
AncSnrGateRS, ANC adjusts the gain of downlink speech signals until the ratio is equal to
AncSnrGateRS.
Based on the actual requirements, the gain of the downlink speech signals ranges from 3 dB to 12 dB. The
AncMaxGain parameter specifies the maximum gain of the downlink speech signals. If the AncMaxGain parameter is
set to a small value, the target ratio of the voice level in the downlink to the noise level in the uplink may be difficult to
achieve. If the AncMaxGain parameter is set to a large value, the user experience may be affected because of the
abrupt change in the downlink voice volume.
If the Tandem Free Operation (TFO) feature is enabled during a call, the ANC feature does not take effect for this call.
Principles
ANR periodically analyzes the speech signals. Through energy estimation, SNR estimation, sound
measurement estimation, and frequency offset estimation of different frequency bands, ANR identifies
the background noise and performs filtering based on decision updates to obtain the time-domain
speech signals after noise restraint.
The ANR feature is enabled in the uplink when the AnrEnFlag parameter is set to UPLINK(On for
uplink).
ANR performs the following operations to restrain the noise:
1. The system performs weighting and window processing on the speech signals.
2. The system performs FFT to convert time-domain speech signals into frequency-domain speech
signals.
3. The system performs energy estimation for frequency-domain speech signals.
4. The system performs the SNR estimation based on the noise energy to obtain the sound
measurement estimation value and the frequency offset estimation value.
5. The signals with a low sound measurement estimation value or a low frequency offset are identified
as the noise signals. The system then calculates the frequency band gain and performs frequency
filtering.
The ANR feature is enabled in the downlink when the AnrEnFlag parameter is set to DOWNLINK(On for dnlink). In
this case, user experience in the background music or ring back tone (RBT) service may be affected.
If the Tandem Free Operation (TFO) feature is enabled during a call, the ANR feature does not take effect for this call.
When using a voice quality test device, for example, the DSLA, you are advised to disable ANR to obtain accurate test
results.
4 VQE3.0
Based on VQE1.0, some optimization is brought into VQE3.0. In VQE3.0, the system can distinguish
music and RBT services from noises more accurately, the function of real-time speech monitoring over
the network is supported, the ACLP feature is added, and the AEC, ALC, ANC, and ANR features are
enhanced. Thus, the user experience is improved. (GBFD-115705 VQE3.0)
A speech channel can be simplified as a module with four ports: source in (Sin), source out (Sout),
remote in (Rin), and remote out (Rout). Figure 4-1 shows the relations between five algorithm modules
of VQE3.0.
Figure 4-1 Relations between five algorithm modules of VQE3.0
CODEC VQE
A
A A A Sout
Sin C
DEC E N L
L
C R C
P
BTS-MS PSTN
A
Rout A A A Rin
C
ENC N L N
L
C C R
P
BSC-TC
The uplink data Sin after decoding is sent to the AEC module, ANR module, ACLP module, and then
ALC module in succession for acoustic echo control, automatic noise restraint, anti-clip processing, and
automatic level control, and finally Sout is obtained.
The downlink data Rin is sent to the ANR module, ACLP module, ALC module, and then ANC module in
succession for automatic noise restraint, anti-clip processing, automatic level control, and automatic
noise compensation. Then, after encoding Rout is obtained.
The ACLP module is behind the AEC module because the anti-clip processing may affect the acoustic
echo control. The ACLP module is before the ALC module because the anti-clip processing may change
the signal level.
4.1 ACLP
Overview
Speech clipping may occur during the electric-acoustic conversion in the MS. Speech clipping leads to
sudden phase change and high-order harmonic wave at sampling points, which will cause noises. As a
result, the user experience is affected.
ACLP is mainly applied in the scenario where the saturation speech clipping occurs during the
electric-acoustic conversion in the MS. In this scenario, the anti-clipping processing is necessary on the
network side. In the ACLP algorithm, the sampling points for input signals are detected to check whether
speech clipping exists. If speech clipping does not exist, no processing is required; if speech clipping
exists, the signals are restored.
Principles
The ACLP feature is enabled in both uplink and downlink when the AclpEnFlag parameter is set to
BOTH(On for uplink and downlink).
The principles of the ACLP feature are as follows:
1. The speech clipping detection is performed to update the speech clipping state factor.
2. The phase balancing filter and the gain control filter are updated according to the speech clipping
state factor.
3. The adaptive filtering is performed on the signal data by using the phase balancing filter and the gain
control filter. In this step, the phase balancing filter relieves the impact of a sudden phase change by
transferring parts of the impact to adjacent sampling points, and the gain adjustment ensures the
wave continuity.
After ACLP is enabled, the speech clipping detection module checks whether speech clipping exists in
real time. If no speech clipping exists, the speech signals are transparently transmitted; if speech
clipping exists, the adaptive filtering is performed for phase balancing. In addition, a state transition
range is set for tradeoff processing if the detection module cannot determine whether speech clipping
exists.
If the Tandem Free Operation (TFO) feature is enabled during a call, the ANR feature does not take effect for this call.
5 Parameters
Table 5-1 Parameters Description
Parameter ID NE MML Description
AecTail BSC6900 SET Meaning: Delay range in which the BSC searches
TCPARA(Optional) for echo and performs echo suppression. If the
actual delay is greater than the value of this
parameter, the BSC does not search for echo or
perform echo suppression.
AncSnrGateRS BSC6900 SET Meaning: The system adjusts the remote input so
TCPARA(Optional) that the ratio of the remote signal to the local
noise is above the value of this parameter.
6 Counters
For the counters, see the BSC6900 GSM Performance Counter Reference.
7 Glossary
For the acronyms, abbreviations, terms, and definitions, see the Glossary.
8 Reference Documents
[1] BSC6900 Feature List
[2] BSC6900 Optional Feature Description
[3] GBSS Reconfiguration Guide
[4] BSC6900 GSM Parameter Reference
[5] BSC6900 GSM MML Command Reference
[6] BSC6900 Performance Counter Reference