3GPP Enhanced Voice

Services (EVS) codec

The Enhanced Voice Services (EVS) codec brings unprecedented quality to voice and
generic audio, including music. Developed by the 3rd Generation Partnership Project
(3GPP), the EVS codec supports the low latency required for real-time communication.
Compared to earlier voice codecs, it can provide much better quality at similar bitrates,
unparalleled quality at higher bitrates, or greatly increased network capacity while
maintaining the same quality level. The EVS codec is fully interoperable with HD voice and
forms a continuous evolution to higher-quality, more natural communication.
This paper discusses the EVS codec, its benefits, and the ways in which it will revolutionize
voice communications. The paper also discusses 3GPP voice and audio codec evolution,
and the role that Nokia has played in developing the EVS codec.

Nokia White Paper

The 3GPP Enhanced Voice Services (EVS) codec
 Contents

Introduction 3
Evolution of 3GPP mobile voice/audio codecs 3
EVS standardization 5
EVS features 5
EVS technology 6
EVS performance 7
Clean speech results 9
Noisy speech results 9
Music and mixed content results 10
Channel condition results 11
Summary of EVS performance 12
EVS capacity benefits for LTE 13
Beyond EVS 15
Conclusion 15
Acronyms 15
References 16

Page 2 www.nokia.com
Introduction
Voice quality plays an important role in the evolution of mobile systems. Users are becoming accustomed
to computer-based voice over IP (VoIP) services that provide extended audio bandwidth. They also expect
high quality from mobile voice services.
High-definition (HD) voice is being rolled out globally, bringing substantial improvement to voice quality in
mobile communications [1]. It allows people to communicate more clearly and easily by using the wider
audio bandwidth (up to 8 kHz) provided by the Adaptive Multi-Rate – Wideband (AMR-WB) codec to make
speech more intelligible and natural, create a feeling of transparent communication, and make speaker
recognition easier [2]. HD voice also doubles the audio bandwidth compared to traditional telephony. It is
therefore not surprising that HD voice makes one-to-one voice calls much more intimate and conference
calls more efficient. The improved intelligibility and natural sound enables clear calls even in noisy
environments.
Although HD voice is still being introduced, the next improvement in voice quality has now emerged [3].
The 3rd Generation Partnership Project (3GPP) recently standardized a new Enhanced Voice Services (EVS)
codec that will take voice quality to a level “beyond HD.” By extending audio bandwidth up to 20 kHz, the
EVS codec covers the full range of human hearing. It is the first 3GPP conversational codec (i.e. with latency
suitable for communication) that provides equally high quality for voice, generic audio such as music, and
content that mixes voice and music.
The EVS codec enables people to use their mobile devices to share music or other content from live events
with quality that creates a realistic impression of the live experience. Developed specifically for IP-based
communications, the EVS codec provides high robustness to delay jitter and packet losses. It reaches the
limits of human audio perception and can cope with network impairments.

Evolution of 3GPP mobile voice/audio codecs

With the new EVS codec, Nokia continues its strong contribution to the development of mobile voice
and audio codecs for GSM and 3GPP systems (3G WCDMA and LTE). Nokia has collaborated with other
companies to develop the following codecs:
• GSM Enhanced Full-Rate (EFR) codec [4], which is used in the 2G GSM system and supports the highest
quality rate in the multi-rate AMR-NB codec
• AMR Narrowband (AMR-NB) codec [5], the default codec for all 3GPP voice services in 3G and beyond
(WCDMA and LTE)
• AMR Wideband (AMR-WB) codec [2], a wideband audio evolution of the AMR-NB codec that provides the
basis for HD voice.
• Extended AMR-WB (AMR-WB+) codec [6], which provides modes for coding stereo signals and supports
generic audio for non-conversational applications such as music streaming
• EVS codec, the next step in 3GPP codec evolution.
Table 1 summarizes the evolution of 3GPP codecs.

Page 3 www.nokia.com
 GSM/GERAN EFR AMR-NB AMR-WB AMR-WB+ EVS
Year standardized 1996 1999 2001 2004 2014
Audio bandwidth Narrowband Narrowband Wideband Fullband Fullband
Super wideband
Wideband
Narrowband
Use in 3GPP Used in the GSM Default codec for Codec for HD Recommended Evolved HD voice
system. voice in 3G and voice. Default codec for generic codec for LTE
beyond (WCDMA, codec for audio in 3G and
LTE) wideband voice beyond (WCDMA,
in 3G and beyond LTE)
(WCDMA, LTE).
Bitrate(s) 12.2 kbit/s 4.75 – 12.2 6.6 – 23.85 6 – 48 kbit/s 5.9 – 128 kbit/s
kbit/s kbit/s

Table 1: 3GPP codecs developed by Nokia in collaboration with other companies

Beyond 3GPP, Nokia has made significant contributions to the following speech codec standards:
• PCS1900 EFR (1995)
• US-TDMA EFR (1996)
• CDMA EVRC (1996)
• ITU-T G.722.2 (2001)
• CDMA VMR-WB (2004)
• ITU-T G.718 (2007) and its super wideband extension (2009).
The evolution of GSM/3GPP voice codecs has significantly improved voice quality and system efficiency,
and played an important role in the success of modern mobile communication. It has continuously
extended audio bandwidth from narrowband voice to the whole range of human hearing, bringing high
quality for speech and generic audio (including music). Robustness to transmission impairments and
system efficiency have improved at the same time.
The new EVS codec substantially improves voice quality, error resilience, and coding efficiency for
narrowband (NB) and wideband (WB) audio bandwidths. The introduction of super wideband (SWB) and
fullband (FB) audio brings further quality enhancements. Compared to HD voice, the EVS codec can
provide substantially improved quality at similar bitrates and unprecedented quality at higher bitrates.
It can also deliver significantly improved network capacity while maintaining the same quality as HD voice.
Figure 1 shows theoretical bandwidths for NB, WB, SWB and FB audio. The actual bandwidths in use may be
somewhat narrower. For example, the bandwidth is typically 300–3400 Hz for NB audio.

Fullband (FB)

Super wideband (SWB)

Wideband (WB)

Narrowband (NB)

0 4000 8000 16000 20000

Frequency (Hz)

Figure 1: Theoretical bandwidths for NB, WB, SWB and FB audio

Page 4 www.nokia.com
One important goal of the EVS codec developers was to enable interaction with the HD voice service.
The EVS codec is fully interoperable with the AMR-WB codec used for HD voice. It represents a
continuous evolution of HD voice to even higher quality and more natural communication. The EVS codec
complements HD voice and provides a solution for achieving higher levels of voice and audio quality.

EVS standardization
Developers began working on the 3GPP EVS codec in 2007 with a pre-study to develop the concept of the
codec. Most EVS specifications were approved in September 2014. Standardization work was completed in
December 2014, for 3GPP Release 12. The EVS codec specifications can be found in [7-18].
The EVS standardization process included three phases:
• Qualification: Pre-selecting the best candidate codecs
• Selection: Choosing the codec that could meet all requirements
• Characterization: Determining the full performance of the new codec.
Codec selection was based on a rigorous process that compared codec performance and other
characteristics against LTE voice codec requirements.
The EVS codec was developed for all-IP 3GPP LTE but will enhance any VoIP or circuit-switched (CS)
system. In 3GPP Release 12, the EVS codec is recommended for voice in packet-switched (PS) multimedia
telephony over LTE and WCDMA. The AMR-WB interoperable mode of EVS is defined as an alternative
implementation of the AMR-WB codec (when the EVS codec is supported), and EVS is defined as the default
codec for SWB and FB speech. Release 13 extends the use of EVS to CS telephony over WCDMA.
The EVS codec was developed jointly by 12 companies: Ericsson, Fraunhofer IIS, Huawei, Nokia, NTT, NTT
DOCOMO, Orange, Panasonic, Qualcomm, Samsung, VoiceAge, and ZTE Corporation.

EVS features
The EVS codec enhances audio quality and improves coding efficiency for NB and WB audio bandwidths
using a wide range of bitrates, starting from 7.2 kbit/s. In addition to fixed-rate coding rates, the codec
supports a source-controlled variable bitrate (SC-VBR) mode at an average bitrate of 5.9 kbit/s for NB
and WB audio. The EVS codec also provides a significant step in voice quality with SWB and FB operation,
starting from 9.6 and 16.4 kbit/s, respectively. It supports a maximum bitrate of 24.4 kbit/s for NB and
128 kbit/s for all other audio bandwidths.
The EVS codec offers input and output sampling at 8, 16, 32, and 48 kHz. To optimize coding quality, an
integrated bandwidth detector automatically adapts to the actual bandwidth of the input signal, which may
be lower than the bandwidth indicated to the codec. The codec’s ability to switch the bitrate at every 20
ms frame allows it to easily adapt to changes in channel capacity.
The codec features discontinuous transmission (DTX) with algorithms for voice and sound activity
detection (VAD) and comfort noise generation (CNG). In the SC-VBR coding mode, DTX and CNG are always
used for inactive speech coding. An advanced error concealment mechanism mitigates the quality impact
of channel errors that result in lost packets. The codec also contains a system for jitter buffer management
(JBM) to tackle the jitter or variation in the delay of received packets. There is also a special channel-aware
mode that can increase robustness in particularly adverse channel conditions. The channel-aware coding
mode operates at 13.2 kbit/s for both WB and SWB audio.

Page 5 www.nokia.com
In addition to the EVS Primary modes described above, the EVS codec supports backward compatibility
with the AMR-WB codec through an interoperable (IO) mode. This IO mode enhances quality compared to
AMR-WB for all nine bitrates between 6.6 kbit/s and 23.85 kbit/s with encoder and decoder improvements,
such as optimized pitch prediction and enhanced post-processing. The EVS codec supports seamless
switching between the AMR-WB IO and EVS Primary modes.
The feature set of the EVS codec results in a highly flexible, dynamically configurable codec that spans
quality ranges from the highest compression rates all the way up to transparent coding. The AMR-WB
IO mode can be used for voice over LTE as an alternative implementation of AMR-WB in terminals and
gateways that support the EVS codec.

EVS technology
The EVS codec uses content-dependent, on-the-fly switching between speech and audio compression
to provide good quality for speech and music signals. It is the first codec to provide core switching
technology at the low algorithmic delay of 32 ms.
Figure 2 presents a high-level view of the structure of the EVS codec.

ENCODER DECODER

EVS modes EVS modes

Jitter buffer management (JBM)

ACELP based BWE ACELP based BWE

POST-PROCESSING
encoder encoder decoder decoder
PRE-PROCESSING

Core, and Core, and

DTX DTX
MDCT based decoder
CHANNEL

switching MDCT based encoder

switching

DTX, CNG encoder DTX, CNG decoder

AMR-WB IO AMR-WB IO
AMR-WB IO mode encoder AMR-WB IO mode decoder

Figure 2: High-level diagram of the EVS codec

Pre-processing includes high-pass filtering at 20 Hz, resampling, signal activity detection, noise update and
estimation, bandwidth detection, and signal analysis and classification. The pre-processing unit’s analysis
and processing capabilities enable a fine-grained classification of the signal type. It applies a targeted
coding method to each case.
Fine tuning of input signal characteristics enables the EVS codec to work with a wide range of content
types. Although signal bandwidth is signaled as an input parameter, the bandwidth detector identifies
lower-than-signaled bandwidth. These capabilities help prevent inefficient coding for band-limited content.
Based on the bitrate and input bandwidth, the EVS codec uses two different internal sampling rates for
the encoding process, thereby allowing the encoded signal to achieve greater fidelity to the original input
signal. Within the EVS modes, core and DTX switching is performed based on information extracted during
pre-processing, as well as on the overall input parameters (bitrate, use of discontinuous transmission,
and input signal bandwidth). If AMR-WB IO mode is selected, encoding is performed according to the
interoperable AMR-WB encoder. The codec can switch between cores and modes at each 20 ms frame
boundary.

Page 6 www.nokia.com
The speech core used in the EVS codec is based on the principles of Algebraic Code-Excited Linear
Prediction (ACELP), inherited from the AMR-WB standard [2]. ACELP relies on the modelling of speech
using linear prediction within an analysis-by-synthesis method. The linear prediction (LP) and excitation
parameters are encoded and most of the bit budget of this model is allotted to the LP parameters. To use
the fine content description of the coder from the preprocessing stage, the codec applies different coding
representations for each signal type. This differentiation requires high memory consumption. The large
range of bitrates that the codec supports further increases the need for differentiation and, implicitly,
memory.
Nokia’s multi-stage structured quantizer technology [19] is based on multiple-scale lattices. It supports
high encoding efficiency that accommodates all signal types, bandwidths, bitrates, and internal sampling
rates while keeping the encoding complexity and read-only memory (ROM) tables within practical limits.
In addition to the flexibility brought by the quantizer structure, controlled alternation between predictive
and non-predictive modes for encoding the LP parameters provides good resilience to frame loss errors.
For speech signals, the part of the bandwidth not covered by the ACELP model for SWB signals is encoded
using a time-domain bandwidth extension (BWE) technique. Multi-bandwidth listening test results show a
significant quality improvement for SWB compared to WB for all supported operating points.
Encoding based on the modified discrete cosine transform (MDCT) is best suited for music and various
background-type signals. Compared with other music-oriented content distribution codecs such as
AAC [20], the EVS codec offers high-quality compression of music signals at low delay and low bitrates.
It delivers these enhancements by using different MDCT-based modes for different content types and
operating modes.
DTX within the EVS modes is important for optimizing battery life in mobile communications. In DTX mode,
CNG replaces transmission of noise in the decoder. Improved VAD helps the codec distinguish between
active speech, active music, and inactive periods (recording noise, background noise), and estimate the
level of the background noise. Based on these decisions, the EVS codec implements two versions of CNG,
one based on LP and the other a frequency domain CNG.
Post-processing tools – such as music enhancer, inactive signal post processing, bass-boost filter, and
formant post filter – help to ensure the high fidelity of the decoded signal. Compared with the AMR-WB
codec, the most notable improvements due to post-processing are audible in noisy channel conditions and
for mixed content.

EVS performance
The EVS codec’s subjective speech and audio quality have been evaluated extensively by means of listening
tests. Several rounds of listening tests were conducted during the 3GPP standardization process. The
most recent rounds were the selection-phase [21] and characterization-phase listening tests [22]. Nokia
has conducted several additional listening tests to better characterize the codec’s subjective speech and
audio quality [23]. 3GPP has published a technical report about the performance characteristics of the EVS
codec [18].
The main attributes that affect subjective speech and audio quality are audio bandwidth, signal type,
channel conditions, and available bitrate.
• Wider audio bandwidth provides additional naturalness to voice quality and enables high quality for
generic audio. Traditionally, NB has been used for telephony. Recently, WB audio using the AMR-WB
codec (HD voice) has gained traction. EVS takes quality to a new level with SWB and FB bandwidths.
Table 2 summarizes the sampling rates and audio bandwidths supported by the EVS codec. The EVS
encoder has a global high-pass filter set at about 20 Hz for all sampling rates. This removes very
low-frequency noise outside the range of human hearing and improves coding.

Page 7 www.nokia.com
 Sampling rate Audio bandwidth
Narrowband (NB) 8 kHz up to 4 kHz
Wideband (WB) 16 kHz up to 8 kHz
Super wideband (SWB) 32 kHz up to 16 kHz
Fullband (FB) 48 kHz up to 20 kHz

Table 2: Sampling rates and audio bandwidths in the EVS codec

• Signal type has a significant effect on coding, especially if the codec is not optimized for speech and
music. For example, the older 3GPP codecs AMR-NB and AMR-WB are optimized primarily for speech
and provide limited audio quality with music signals, even at the highest available bitrates. By contrast,
EVS is designed to support speech and music, and to be signal independent. It provides reasonable
quality for generic audio even at low bitrates. Despite its name, EVS is much more than a voice codec.
• Channel conditions refer to the real-time transmission channel (e.g. LTE radio network) where,
depending on the radio channel characteristics and network conditions (e.g. congestion), some audio
packets may be lost, corrupted or received too late for real-time decoding. In these cases, the audio
codec has to perform frame error concealment (FEC). Instead of muting or producing disturbing
sounds, the decoder uses special algorithms to generate an artificial but pleasant-sounding signal to
replace the lost signal frames. Speech codecs can typically handle a small number of lost frames (up to
about 3 percent) without severe artefacts. The EVS codec improves these handling capabilities. Even
with frame loss of more than 10 percent, EVS provides understandable voice output to temporarily
cope with such extreme conditions without severe artefacts.
• The bitrate influences network and radio resource consumption. The most typical bitrates in 3GPP voice
telephony are around 12–13 kbit/s. AMR-NB supports bitrates from 4.75–12.2 kbit/s. AMR-WB supports
bitrates from 6.6–23.85 kbit/s. EVS supports a wide range of bitrates from 5.9–128 kbit/s. Subjective
quality should improve with increasing bitrates. Depending on which modes are used, EVS can provide
significantly improved quality over AMR-NB and AMR-WB at similar bitrates, quality raised to an
unprecedented level at increased bitrates, or significantly improved network capacity while maintaining
the same quality.
Figures 3–8 contain results from subjective listening tests conducted by Nokia. These results are
represented on a mean opinion score (MOS) scale. The MOS is calculated based on how non-expert
listeners cast votes for different test signal segments containing speech sentences, noisy speech excerpts,
or music sequences. In a typical single listening test, 24–32 different listeners hear each condition. Each
listener casts votes for 4–8 different signal segments with the same conditions. The individual MOS is
typically an average of 96–256 casted votes.
The quality scale in the standardized P.800 test is from 1–5 [31]. The Nokia tests extended this scale so
that end points could be scored from 1 (poor) to 9 (excellent). The range extension gives better resolution
in the results. The tests used Absolute Category Rating (ACR) with a nine-point scale, also known as the
ACR9 MOS methodology.
The tests covered several other standardized codecs and reference conditions to provide insight on the
level of improvement delivered by the EVS codec. The reference conditions and codecs included:
• Direct reference conditions with limited audio bandwidth but no coding. The tests used the following
low-pass cut-off frequencies: 20 kHz for FB, 16 kHz for SWB, 8 kHz for WB, and 4 kHz for NB.
• AMR-NB codec, also known as AMR [25]
• AMR-WB wideband codec [26]

Page 8 www.nokia.com
• ITU-T G.722.1 Annex C [27]. This low-complexity SWB voice codec has an audio bandwidth of 14 kHz. It
is used in teleconferencing systems.
• ITU-T G.718 Annex B [28]. This is an embedded (8–64 kbit/s) speech codec for NB, WB and SWB audio.
The tests used SWB audio.
• ITU-T G.719 [29], a FB voice and audio codec used in teleconferencing systems
• IETF Opus, version opus-tools-0.1.9-win32.zip [30]. The tests used the constant bitrate (CBR)
configuration.

Clean speech results

The clean speech (i.e. no background noise) results in Figure 3 show that EVS delivers significantly better
quality than AMR or AMR-WB at all operating points.
FB
8 SWB

7 G.722.1C

A WB G.718B

C 6 G.719
R
Opus CBR
9
5
AMR
M
NB AMR-WB
O 4
S EVS-NB

3 EVS-WB

EVS-
SWB/FB
2
0 8 16 24 32 kbit/s

Figure 3: Clean speech MOS values

In the SWB and FB modes, EVS delivers significantly better quality than G.718B, G.722.1C, or G.719. At
bitrates below 32 kbit/s, EVS in the SWB and FB modes delivers better quality than Opus CBR, with the
quality difference becoming substantial below 24.4 kbit/s. Notably, EVS-SWB 9.6 kbit/s is superior to
AMR-WB 23.85 kbit/s and Opus CBR 20 kbit/s, providing better voice quality at less than half the bitrate.

Noisy speech results

The noisy speech (i.e. speech with background noise) results in Figure 4 are very similar to the clean speech
results. The exception is that G.722.1C and G.719 perform somewhat better in noisy speech conditions
than the other codecs in the test. This phenomenon is well known from previous listening tests [24].
Nevertheless, EVS in the SWB and FB modes delivers significantly better quality than either of these other
two codecs. The test used office, street, car, and cafeteria noise types.

Page 9 www.nokia.com
 8 FB
SWB

7 G.722.1C

A G.718B
WB
C 6 Opus CBR
R
9 AMR
5
AMR-WB
M
NB EVS-NB
O 4
S EVS-WB

3 G.719

EVS-
SWB/FB
2
0 8 16 24 32 kbit/s

Figure 4: Noisy speech MOS values

Music and mixed content results

The music and mixed content results in Figure 5 are quite similar to the clean speech results. For Opus,
however, the lack of quality improvement at increased bitrates is more visible at 20 kbit/s. With music and
mixed content, EVS shows excellent performance compared to Opus, AMR, and AMR-WB. At each bitrate in
the range 5.9–24.4 kbit/s, EVS achieves an MOS up to two points better than any other codec. The reason
for this good performance at low bitrates is the switched ACELP/transform coding cores used in EVS. The
content used in the test intermixed speech and music signals (e.g. radio advertisements and telephone
on-hold messages/music).
8
FB
SWB

7 G.722.1C
WB
A G.718B

C 6 G.719
R
9 Opus CBR
5
NB AMR
M
O 4 AMR-WB

S EVS-NB

3 EVS-WB

EVS-
SWB/FB
2
0 8 16 24 32 kbit/s

Figure 5: Music and mixed content MOS values

Page 10 www.nokia.com
Channel condition results
Figures 6 and 7 show how EVS performs with varying amounts of channel errors. The frame error rate (FER)
is shown as a percentage on the X-axis and the MOS is shown on the Y-axis. Direct NB, WB, and FB signals
obtained MOSs of 4.91, 7.04, and 8.00 in this listening test. The tests used all signal types: clean speech,
noisy speech, music, and mixed content. The channel-aware mode of the EVS codec was not used in these
tests.
Figure 6 shows results for AMR, AMR-WB and EVS at bitrates of 12–13 kbit/s (13.2 kbit/s for EVS, 12.65
kbit/s for AMR-WB and 12.2 kbit/s for AMR). Comparisons of EVS-SWB and EVS-WB to AMR-WB and
EVS-NB to AMR reveal the inherent robustness of the EVS codec. EVS shows significant improvement at all
operating points. The performance improvement grows as the FER increases.
7

A
C 5
AMR - 12,2
R
EVS-NB - 13,2
9
4
AMR-WB - 12,65

M EVS-WB - 13,2

O 3 EVS-SWB - 13,2

S
2

1
0 3 6 10 15 FER%

Figure 6: Audio quality when channel errors are present – EVS, AMR-WB, and AMR codecs

EVS-SWB, EVS-WB, and EVS-NB show little or no quality degradation at a 3 percent FER when compared to
a clean channel. Even at a 4–6 percent FER, EVS-WB and EVS-SWB provide quality comparable to AMR-WB
without errors, and EVS-NB provides quality comparable to AMR without errors. At a 10 percent FER,
EVS-WB and EVS-SWB provide usable audio quality (similar to AMR-WB at a 4–5 percent FER) as a
temporary measure to cope with errors, thus doubling the frame error robustness for similar quality. At
the highest measured FER of 15 percent, EVS-WB provides better quality than AMR at a 3 percent FER. At
high error rates above 8 percent, FER EVS-NB provides higher audio quality than AMR-WB, although it is
limited to a narrower signal bandwidth.
Figure 7 shows how EVS performs compared to the highest bitrate of AMR-WB (23.85 kbit/s). The inherent
robustness of EVS is evident from these results. EVS-WB and EVS-SWB at 13.2 kbit/s both perform
significantly better than AMR-WB at 23.85 kbit/s for all FERs, even though the bitrate is almost halved.
Comparing EVS-SWB at 32 kbit/s to AMR-WB at 23.85 kbit/s reveals an average improvement of 1.4 MOS
points across all operating points. Even at a 6 percent FER, EVS-SWB at 32 kbit/s provides significantly
better quality than AMR-WB at 23.85 kbit/s in clean channel conditions. EVS-SWB at 13.2 kbit/s with a 6
percent FER provides similar quality to AMR-WB at 23.85 kbit/s in clean channel conditions. EVS-FB at 48
kbit/s provides unprecedented quality for communications at all operating points, averaging a 1.8 MOS
point improvement over AMR-WB at 23.85 kbit/s.

Page 11 www.nokia.com
 9

7
A
C
6 AMR-WB - 23,85
R
EVS-WB - 13,2
9
5
EVS-SWB - 13,2

M 4 EVS-SWB - 32

O EVS-FB - 48

S 3

1
0 3 6 10 15 FER%

Figure 7: Audio quality when channel errors are present – EVS codec compared to the highest-bitrate mode
of AMR-WB at 23.85 kbit/s

Summary of EVS performance

Figure 8 combines all the Nokia test results in a single graph. The graph also includes results for higher
bitrates (up to 128 kbit/s). The combined results show that the quality delivered by EVS is better than,
or statistically equivalent to, that of any other codec at all bitrates. EVS-NB at 24.4 kbit/s is statistically
equivalent to NB direct. EVS-WB at 32 kbit/s is statistically equivalent to wideband direct. In addition,
EVS-SWB and EVS-FB reach statistical equivalence to direct at 64 kbit/s and 96 kbit/s, respectively.

8 FB
SWB

A WB

C 6
G.722.1C
R
G.718B
9
5 G.719

M Opus CBR
NB
O 4 AMR
S AMR-WB
EVS-NB
3
EVS-WB
EVS-SWB/FB
2
0 8 16 24 32 40 48 64 96 128 kbit/s

Figure 8: Combined MOS values

Page 12 www.nokia.com
The results show that the 3GPP EVS codec produces cutting-edge voice and audio quality across all tested
bitrates and bandwidths. The codec excels at bitrates up to 32 kbit/s, which are of utmost importance for
the deployment of cost-effective mobile services. Compared to Opus and other tested codecs, EVS can
provide the same or better quality at about half the bitrate. For example, EVS-SWB at 9.6 kbit/s delivers
significantly better quality than Opus CBR at 20 kbit/s. EVS-SWB at 16.4 kbit/s provides the same quality
as G.722.1C at 48 kbit/s.
EVS-SWB at 13.2 kbit/s provides an MOS of 6.15, AMR-WB at 12.65 kbit/s provides an MOS of 4.95 and
AMR at 12.2 kbit/s provides an MOS of 3.51. This means that the improvement from AMR-WB to EVS-SWB
(1.2 points) is almost as large as the improvement from AMR to AMR-WB (1.44 points). At slightly higher
operating points – EVS-SWB at 24.4 kbit/s versus AMR-WB at 23.85 kbit/s – the improvement is almost 1.7
MOS points. This is a larger improvement than that provided between AMR at 12.2 kbit/s and AMR-WB at
12.65 kbit/s.
The SWB and FB modes of EVS at the same bitrates show similar performance. They are represented by a
single curve in Figure 8. For audio with frequencies above 16 kHz (and for listeners capable of hearing such
high frequencies), the FB mode provides significant quality improvement over the SWB mode.

EVS capacity benefits for LTE

Nokia has studied the voice capacity impact of using the EVS codec in a 3GPP LTE network. Based on the
results of the subjective listening tests presented in the previous section, Nokia analyzed the EVS audio
bandwidths and bitrates that produce subjective quality equal to or better than the most relevant use
cases of the AMR and AMR-WB codecs. The most relevant reference use cases are:
• AMR 12.2 kbit/s. This codec bitrate is used for cases that require NB communications– for example,
when interworking with a CS network that supports only NB voice.
• AMR-WB 12.65 kbit/s. This WB codec bitrate can be used when optimal interworking with HD voice
towards a 3GPP 2G or 3G network is desired. This bitrate ensures optimal voice quality by enabling
transcoder-free operation (TrFO) between LTE and CS networks.
• AMR-WB 23.85 kbit/s. This bitrate provides the best voice quality with the AMR-WB codec.
Lower AMR/AMR-WB bitrates can also be used if rate control is active on the CS side.
Table 3 shows the EVS audio bandwidths and bitrates that produce subjective quality equal to or better
than that produced by AMR or AMR-WB for typical conversational voice scenarios, including clean speech
and speech with background noise.

Reference Equal bandwidth Wider bandwidth

AMR 12.2 kbit/s EVS-NB 8.0 kbit/s EVS-WB 5.9 kbit/s
AMR-WB 12.65 kbit/s EVS-WB 9.6 kbit/s EVS-SWB 9.6 kbit/s
AMR-WB 23.85 kbit/s EVS-WB 13.2 kbit/s EVS-SWB 9.6 kbit/s

Table 3: EVS bandwidths and bitrates that equal or exceed subjective quality of reference scenarios

EVS voice capacity estimations in LTE are based on existing LTE voice capacity simulations for the AMR
codec using 5.9 kbit/s and 12.2 kbit/s as reference points [32]. The main system simulation parameters
were aligned with [33]. The reference AMR capacities were 410 and 240 users for 5.9 kbit/s and 12.2
kbit/s, respectively, at 5 MHz system bandwidth in the uplink direction and using a semi-persistent
scheduling scheme. The uplink direction was chosen because LTE voice capacity is uplink limited and thus

Page 13 www.nokia.com
defines the end-to-end system capacity. Better voice capacity can be achieved with an interface that
supports semi-persistent scheduling over the air rather than dynamic scheduling.
The voice capacities of AMR-WB 23.85 kbit/s and EVS bitrates were interpolated or extrapolated from the
capacities of the reference points. A regression analysis tool was used to perform the interpolation and
extrapolation.
Figure 9 illustrates estimated EVS capacity gains compared to the AMR and AMR-WB reference cases
shown in Table 3. The estimates assume that no other user plane or layer 1 control channel factor limits
these gains.

200%

180%

160%

140%

120%

100% EVS with the same bandwidth

EVS with wider bandwidth
80%
EVS-WB 13.2
EVS-SWB 9.6

EVS-SWB 9.6
EVS-WB 5.9

EVS-WB 9.6
EVS-NB 8.0

60%

40%

20%

0%
EVS vs. AMR 12.2 EVS vs. AMR-WB 12.65 EVS vs. AMR-WB 23.85

Figure 9: EVS voice capacity compared to AMR and AMR-WB.

The highest capacity gain is achieved when the EVS-SWB 9.6 kbit/s codec is used instead of the AMR-WB
23.85 kbit/s codec. A near doubling of capacity is possible. Significant capacity gains are also achievable
against the NB reference scenario, where the EVS-WB 5.9 kbit/s codec provides a capacity improvement of
about 70 percent over the AMR 12.2 kbit/s codec.
The lowest expected capacity gain is generated when the EVS codec is used instead of the AMR-WB 12.65
kbit/s codec. In this case, the EVS-WB 9.6 kbit/s and EVS-SWB 9.6 kbit/s codecs both provide a capacity
gain of about 20 percent. The capacity gain is the same for the WB and SWB modes because 9.6 kbit/s is
the lowest available bitrate for the EVS-SWB mode. The reduced resource consumption of VoLTE on the LTE
data channel provides additional capacity for other services or applications.
At low bitrates, the EVS codec improves VoLTE uplink coverage in coverage-limited cells while maintaining
the same voice quality at the cell edge. This improvement is created because the LTE terminal can focus
its power on a smaller number of physical resource blocks. As shown in figures 6 and 7, the EVS codec can
provide comparable audio quality at significantly higher FER levels than the AMR or AMR-WB codecs. It also
improves uplink coverage. The EVS-WB/SWB 13.2 kbit/s codec with the channel-aware mode is well suited
for use in large LTE cells. Its significantly higher error resilience decompensates sporadic packet losses,
which typically arise as the terminal approaches end of coverage.

Page 14 www.nokia.com
Beyond EVS
The desire to improve naturalness of voice communications and enhance robustness against transmission
errors has been the main driver for new voice codecs. Natural voice quality is particularly important
for multi-party conference calls and telepresence applications, which benefit from improved speaker
recognition. It is also important for preventing listener fatigue during long sessions. The EVS codec
currently supports coding of stereo signals, but only by means of coding two separate mono channels.
Extending EVS in future 3GPP releases to stereo and multi-channel coding (with spatial localization of the
participants) will be a natural evolution of EVS. It will bring further improvements to the user experience
and to system efficiency.

Conclusion
3GPP has standardized a new codec for EVS. The EVS codec is the successor to the HD mobile voice codec
AMR-WB. It provides full interoperability with HD voice. EVS offers cutting-edge performance that exceeds
the capabilities of other codec standards from ITU-T, IETF and 3GPP. It excels with high audio quality for
speech and music, outstanding compression efficiency, and superior robustness against transmission
impairments. These characteristics are all essential for mobile operators seeking to deliver cost-effective
mobile services and achieve long-term business success. Compared to HD voice, the EVS codec provides
substantially improved quality at similar bitrates and unprecedented quality at higher bitrates. It also
significantly improves network capacity while maintaining the same quality as HD voice.

Acronyms
2G 2nd Generation
3GPP The 3rd Generation Partnership Project
AAC Advanced Audio Coding
ACELP Algebraic Code-Excited Linear Prediction
ACR Absolute Category Rating
ACR9 Absolute Category Rating with 9-point scale
AMR Adaptive Multi-Rate (same codec as AMR-NB)
AMR-NB Adaptive Multi-Rate – Narrowband
AMR-WB Adaptive Multi-Rate – Wideband
AMR-WB+ Extended AMR-WB
BWE bandwidth extension
CBR constant bitrate
CDMA code division multiple access
CNG comfort noise generation
CS circuit-switched
DTX discontinuous transmission
EDGE Enhanced Data rates for GSM Evolution
EFR Enhanced Full-Rate
EVRC Enhanced Variable Rate Codec
EVS Enhanced Voice Services

Page 15 www.nokia.com
FB fullband
FEC frame error concealment
FER frame error rate
GERAN GSM EDGE Radio Access Network
GSM Global System for Mobile communications
HD high definition
IETF Internet Engineering Task Force
IO interoperable
ITU-T International Telecommunication Union - Telecommunication standardization sector
JBM jitter buffer management
LP linear prediction
LTE Long Term Evolution
MDCT modified discrete cosine transform
MOS mean opinion score
NB narrowband
PCS Personal Communications Service
PS packet-switched
SC-VBR source-controlled variable bitrate
SWB super wideband
TrFO transcoder-free operation
US-TDMA United States – Time Division Multiple Access
VAD voice/sound activity detection
VMR-WB Variable-Rate Multimode Wideband
VoIP voice over Internet Protocol
VoLTE voice over LTE
WB wideband
WCDMA Wideband Code Division Multiple Access

References
[1] http://www.gsacom.com/hdvoice/
[2] B. Bessette et al., “The Adaptive Multi-Rate Wideband Speech Codec (AMR-WB)”, IEEE Transactions
on Speech and Audio Signal Processing, Vol. 10, No. 8, November 2002.
[3] S. Bruhn et al., “Standardization of the new 3GPP EVS Codec”, IEEE International Conference on
Acoustics, Speech and Signal Processing (ICASSP) 2015, Brisbane, Australia, April 2015.
[4] K. Järvinen et al, “GSM Enhanced Full Rate Codec”, Proc. of IEEE International Conference on
Acoustics, Speech and Signal Processing, Munich, Germany, April 1997.
[5] K. Järvinen, “Standardisation of the Adaptive Multi-Rate Codec”, Proc. of X European Signal
Processing Conference (EUSIPCO), Tampere, Finland, September 2000.
[6] J. Mäkinen et al. “AMR-WB+: a new audio coding standard for 3rd generation mobile audio services”,
Proc. of IEEE International Conference on Acoustics, Speech and Signal Processing, Philadelphia,
USA, March 2005.

Page 16 www.nokia.com
[7] 3GPP TS 26.441, “Codec for Enhanced Voice Services (EVS); General overview”.
Online: http://www.3gpp.org/DynaReport/26441.htm
[8] 3GPP TS 26.442, “Codec for Enhanced Voice Services (EVS); ANSI C code (fixed-point)”.
Online: http://www.3gpp.org/DynaReport/26442.htm
[9] 3GPP TS 26.443, “Codec for Enhanced Voice Services (EVS); ANSI C code (floating-point)”.
Online: http://www.3gpp.org/DynaReport/26443.htm
[10] 3GPP TS 26.444, “Codec for Enhanced Voice Services (EVS); Test sequences”.
Online: http://www.3gpp.org/DynaReport/26444.htm
[11] 3GPP TS 26.445, “Codec for Enhanced Voice Services (EVS); Detailed algorithmic description”.
Online: http://www.3gpp.org/DynaReport/26445.htm
[12] 3GPP TS 26.446, “Codec for Enhanced Voice Services (EVS); Adaptive Multi-Rate - Wideband
(AMR-WB) backward compatible functions”. Online: http://www.3gpp.org/DynaReport/26446.htm
[13] 3GPP TS 26.447, “Codec for Enhanced Voice Services (EVS); Error concealment of lost packets”.
Online: http://www.3gpp.org/DynaReport/26447.htm
[14] 3GPP TS 26.448, “Codec for Enhanced Voice Services (EVS); Jitter buffer management”.
Online: http://www.3gpp.org/DynaReport/26448.htm
[15] 3GPP TS 26.449, “Codec for Enhanced Voice Services (EVS); Comfort Noise Generation (CNG)
aspects”. Online: http://www.3gpp.org/DynaReport/26449.htm
[16] 3GPP TS 26.450, “Codec for Enhanced Voice Services (EVS); Discontinuous Transmission (DTX)”.
Online: http://www.3gpp.org/DynaReport/26450.htm
[17] 3GPP TS 26.451, “Codec for Enhanced Voice Services (EVS); Voice Activity Detection (VAD)”.
Online: http://www.3gpp.org/DynaReport/26451.htm
[18] 3GPP TR 26.952, “Codec for Enhanced Voice Services (EVS); Performance Characterization”.
Online: http://www.3gpp.org/DynaReport/26952.htm
[19] A. Vasilache et al., “Flexible spectrum coding in the 3GPP EVS codec”, IEEE International Conference
on Acoustics, Speech and Signal Processing (ICASSP) 2015, Brisbane, Australia, April 2015.
[20] M. Bosi et al., “ISO/IEC MPEG-2 advanced audio coding”, Journal of the Audio engineering society
45.10, pp. 789-814, 1997.
[21] 3GPP Tdoc S4-141065, “Report of the Global Analysis Lab for EVS Selection Phase”, August 2014.
Online: http://www.3gpp.org/ftp/tsg_sa/WG4_CODEC/TSGS4_80bis/ Docs/S4-141065.zip
[22] 3GPP Tdoc S4-141161, “Report of the Global Analysis Lab for the EVS Characterization Phase”,
3GPP, November 2014.
Online: http://www.3gpp.org/ftp/tsg_sa/WG4_CODEC/TSGS4_81/Docs/ S4-141161.zip
[23] A. Rämö and H. Toukomaa, “Subjective Quality Evaluation of the 3GPP EVS Codec”, IEEE
International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2015, Brisbane,
Australia, April 2015.
[24] A. Rämö and H. Toukomaa, “Voice quality characterization of IETF Opus codec”, in Proc. Interspeech,
pp. 2541–2544, Florence, Italy, August 2011.
[25] 3GPP TS 26.090, “Adaptive multi-rate (AMR) speech codec; Transcoding functions”.
Online: http://www.3gpp.org/DynaReport/26090.htm
[26] 3GPP TS 26.190, “Adaptive multi-rate wideband (AMR-WB) speech codec; Transcoding functions”.
Online: http://www.3gpp.org/DynaReport/26190.htm
[27] ITU-T G.722.1, “Low-complexity coding at 24 and 32 kbit/s for handsfree operation in systems with
low frame loss”, ITU, May 2005. Online: http://www.itu.int/rec/T-REC-G.722.1/en

Page 17 www.nokia.com
[28] ITU-T G.718, “Amendment 2, Frame error robust narrow-band and wideband embedded
variable bit-rate coding of speech and audio from 8–32 kbit/s; Amendment 2: New Annex B on
superwideband scalable extension for ITU-T G.718 and corrections to main body fixed-point C-code
and description text”, ITU, March 2010. Online: http://www.itu.int/rec/T-REC-G.718/en
[29] ITU-T G.719, “Low-complexity, full-band audio coding for high-quality, conversational applications”,
ITU, June 2008. Online: http://www.itu.int/rec/T-REC-G.719/en
[30] J.-M. Valin et al., “Definition of the Opus audio codec”, IETF RFC 6716, September 2012.
Online: http://opus-codec.org/
[31] ITU-T P.800, “Methods for subjective determination of transmission quality”, ITU, August 1996.
Online: https://www.itu.int/rec/T-REC-P.800-199608-I/en
[32] H. Holma and A. Toskala, “LTE for UMTS: Evolution to LTE-Advanced”, Second Edition, John Wiley &
Sons, 2011.
[33] NGMN Alliance, “Next Generation Mobile Networks Radio Access Performance Evaluation
Methodology”, January 2008. Online: http://www.ngmn.org/uploads/media/NGMN_Radio_Access_
Performance_Evaluation_Methodology.pdf

Page 18 www.nokia.com
Nokia is a registered trademark of Nokia Corporation. Other product and company names mentioned herein may be trademarks or trade names of their respective
owners.

Nokia Oyj
Karaportti 3
FI-02610 Espoo
Finland
Tel. +358 (0) 10 44 88 000

Product code SR1702006833EN

© Nokia 2017.

Page 19 www.nokia.com