Study on Speech Transmission under Varying QoS Parameters

in a OFDM Communication System

Monika Zamlynska

, Przemyslaw Falkowski-Gilski

, Grzegorz Debita

and Bogdan Miedzinski

Faculty of Management, General Tadeusz Kosciuszko Military University of Land Forces,

Czajkowskiego 109, Wroclaw, Poland

Faculty of Electronics, Telecommunications and Informatics, Gdansk University of Technology,

Narutowicza 11/12, Gdansk, Poland

Faculty of Electrical Engineering, Wroclaw University of Science and Technology,

Wybrzeze Wyspianskiego 27, Wroclaw, Poland

Keywords: Coding, Communication Applications, Compression, Speech Processing.

Abstract: Although there has been an outbreak of multiple multimedia platforms worldwide, speech communication is

still the most essential and important type of service. With the spoken word we can exchange ideas, provide

descriptive information, as well as aid to another person. As the amount of available bandwidth continues to

shrink, researchers focus on novel types of transmission, based most often on multi-valued modulations,

multiple channels and related sub-carriers. Currently, OFDM (Orthogonal Frequency Division Multiplexing)

is widely utilized both in wired and wireless transmission. It includes terrestrial and online digital services,

such as cellular systems and broadcasting standards. This paper is focused on varying QoS (Quality of

Service) aspects, related with the OFDM telecommunication system, with respect to speech signals.

It involves a group of four language sets, namely: American English, British English, German, and Polish.

Results of this study may aid both researchers and professionals involved in designing everyday

communication services as well as supplementary back-up services.

1 INTRODUCTION

Designing and maintaining reliable communication

services is a challenging task. With the outbreak of

both desktop and mobile devices, followed by novel

modulation and coding schemes, user expectations

regarding the level of quality of a particular service

continues to grow (Hossfeld et al., 2014; Boz et al.,

2019; Kostek, 2019; Falkowski-Gilski and Uhl,

2020). Currently, more and more solutions are based

on OFDM (Orthogonal Frequency Division

Multiplexing) (Cioni, Corazza, Neri and Vanelli-

Coralli, 2006; Aragón-Zavala, Angueira, Montalban

and Vargas-Rosales, 2021). With fluctuating

bandwidth conditions in heterogeneous networks, it is

important to study how does it affect the quality of

transmission.

https://orcid.org/0000-0001-8920-6969

https://orcid.org/0000-0003-1984-4740

https://orcid.org/0000-0001-5354-7024

2 ABOUT THE STUDY

The study involved a set of speech samples, coded

and then processed in our custom-build OFDM

telecommunication system, with respect to QoS

(Quality of Service) requirements.

2.1 QoS Requirements

According to the 3GPP (3rd Generation Partnership

Project), when examining digital voice

communication services, they can be divided into

three groups (3GPP, 2011). Table 1 describes

principle voice (speech) services with their

requirements, including delay and BER (Bit Error

Rate).

Zamlynska, M., Falkowski-Gilski, P., Debita, G. and Miedzinski, B.

Study on Speech Transmission under Varying QoS Parameters in a OFDM Communication System.

DOI: 10.5220/0010719800003058

In Proceedings of the 17th International Conference on Web Information Systems and Technologies (WEBIST 2021), pages 641-646

ISBN: 978-989-758-536-4; ISSN: 2184-3252

641

Table 1: Principle voice communication services with

QoS requirements.

Service Delay [ms] BER

Conversational voice

(real-time transmission)

100 10

-2

Non-conversational voice

(buffered streaming)

300 10

-6

Interactive voice

(live streaming)

100 10

-3

This particular study is focused on speech (voice)

communication. According to Table 1, these services

require a delay from less than 100 ms (conversational

voice) to less than 300 ms (non-conversational voice).

Whereas, when it comes to error rate, the accepted

threshold ranges from 10

-2

or 10

-3

up to 10

-6

depending on the variant. This of course can affect the

quality of transmitted speech samples.

2.2 Speech Signal Samples

The signal samples used during study were sourced

from ITU-T P.501 (ITU, 2017), and consisted of

sentences spoken by both female and male lectors in

different languages. When examining the

international profile of the broadcasting and

streaming industry, we have selected 4 language sets:

American English (AE), British English (EN),

German (GE), and Polish (PL), as described in

Table 2.

These samples were originally available in the

WAV 16-bit PCM format. Next, each sample was

coded using the Ogg Vorbis format (Kosaka et al.,

2002; Kosaka, Okuhata, Onoye and Shirakawa,

2005), the bitrate was set to 32 kbps, whereas the

initial sampling frequency was changed to 44.1 kHz.

Then all of them were transmitted via our

communication system.

3 SIMULATED

COMMUNICATION SYSTEM

In order to assess the quality of transmission in the

modeled telecommunication system, we have utilized

the Matlab/Simulink environment. The simulated

model enabled to determine the BER for the

transmission of audio files. The block diagram is

shown in Figure 1.

Table 2: Tested speech signal samples.

Filename S

oken sentence

AEfemale1 We need grey to keep our mood healthy.

Pack the records in a neat thin case.

AEfemale2 The stems of the tall glasses cracked and

broke.

The wall phone rang loud and often.

AEmale1 The shelves were bare of both jam or

crackers.

to ever

child is the swan boat.

AEmale2 Both brothers were the same size.

In some form or other we need fun.

ENfemale1 These days a chicken leg is a rare dish.

The hogs were fed with chopped corn

and

arba

ENfemale2 Rice is often served in round bowls.

A large size in stockings is hard to sell.

ENmale1 The juice of lemons makes fine punch.

Four hours of stead

work faced us.

ENmale2 The birch canoe slid on smooth planks.

Glue the sheet to the dark blue

ackground.

GEfemale1 Zarter Blumenduft erfüllt den Saal.

Wisch den Tisch doch s

äter ab.

GEfemale2 Sekunden entscheiden über Leben.

Flieder lockt nicht nur die Bienen.

GEmale1 Gegen Dummheit ist kein Kraut

gewachsen.

Alles wurde wieder ab

esa

GEmale2 Überquere die Strasse vorsichtig.

Die drei Männer sind begeistert.

PLfemale1 Pielęgniarki były cierpliwe.

Przebiegał szybko przez ulicę.

PLfemale2 Ona była jego sekretarką od lat.

Dzieci często

łaczą kied

są

łodne.

PLmale1 On był czarującą osobą.

Lato wreszcie nadeszło.

PLmale2 Większość dróg było niezmiernie

zatłoczonych.

Mam

bardzo entuz

ast

czn

zes

ół.

The modeled communication link was based on a

well-known broadcasting chain utilized e.g. in digital

audio broadcasting (DAB) (Harada and Prasad, 2002;

Kim, Lee, Park and Lee, 2014; Zhang, Wang, Wang

and Lu, 2017). The principle parameters are described

in Table 3.

Figure 1: Block diagram of the evaluated speech

communication system.

QQSS 2021 - Special Session on Quality of Service and Quality of Experience in Systems and Services

642

The audio content used during the simulation was

encoded using the Ogg Vorbis codec. It was

transmitted over the channel in a single channel

(mono) mode. Therefore, it did not require

multiplexing during transmission. The coding itself

was based on the DAB system.

Table 3: Principle parameters of the simulated speech

communication system.

Paramete

Value

Sampling frequency 44.1 kHz

Audio mode Mono

Bitrate 32 kb

Audio codec O

Vorbis

Coding rate 3/4

Protection coding FEC 1/4 (BCH and

convolutional codin

Modulation DQPSK and OFDM

Number of sub-carriers 1350

Carrier spacing 1 kHz

Channel width 1.44 MHz

Channel model AWGN

The pre-encoded signal was BCH

(Bose-Chaudhuri-Hocquenghem) encoded, which is

a subclass of cyclic error correction codes. BCH is a

well-known and widely-used coding scheme. It is

quite simple, flexible and energy-efficient (Agarwal

and Patra, 2012). The frame in BCH is of fixed length,

in order to ensure both performance and speed.

The second codec was responsible for

convolutional encoding, as recommended in DAB.

It encodes a sequence of binary input vectors in order

to create a sequence of binary output vectors

(Clark Jr. and Cain, 1981). It can process multiple

symbols simultaneously. According to this standard

(ETSI, 2001), octal polynomials: 133, 171, 145, 133,

were used with a constant length of 7.

Later on, data was interleaved, according to a

random permutation. In this step, all data were

converted into symbols by means of DQPSK

(Differential Quadrature Phase Shift Keying)

baseband modulation combined with OFDM.

OFDM maintains the orthogonality of the

sub-carriers, which reduces the risk of interference.

This scheme transforms one transmission into several

data streams that are less prone to corruption.

The signal was split into N parallel streams.

Later on the IDFT (Inverse Discrete Fourier

Transform) was calculated for each symbol. In case

of the OFDM demodulator, a reverse operation was

performed. The sub-carried symbol was transformed

by the DFT algorithm.

4 RESULTS

In this experiment, the Ogg Vorbis audio files,

with a bitrate of 32 kbps and a sampling frequency of

44.1 kHz, were fed into the simulator. The input was

composed of four language sets, namely: American

English (AE), British English (EN), German (GE),

and Polish (PL). Furthermore, in each case both

female and male voices were distinguished.

All files were processed one by one, in a queue.

The initial (input) as well as resulting (output) format

remained unaltered. Additionally, the content

processing chain did not cause change in either bitrate

nor sampling frequency.

Overall, the test involved 24 audio files, with

signal-to-noise ratio (SNR) ranging from 2.66 to 4.8

dB. The BER was monitored both before forward error

correction (FEC) protection coding (so-called channel

BER) and after FEC protection coding. Results of this

experiment are described in Tables 4-7.

Results for the American English (AE) language

set are described in Table 4. Whereas, results for the

following sets of speech samples, that is:

British English (EN), German (GE), and Polish (PL),

are described in Table 5, 6, and 7, respectively.

Table 4: Results for transmitted speech samples in

American English.

File name SNR

[dB]

BER

after FEC

BER

before FEC

AEfemale1 2.660 1.1088E-02 1.5166E-01

3.530 1.0507E-03 1.2295E-01

4.080 1.1567E-04 1.0594E-01

4.450 1.0092E-05 9.4984E-02

4.745 1.1755E-06 8.6652E-02

AEfemale2 2.660 1.0806E-02 1.5184E-01

3.530 1.0295E-03 1.2319E-01

4.080 1.0437E-04 1.0602E-01

4.450 1.0999E-05 9.5014E-02

4.745 1.2548E-06 8.6657E-02

AEmale1 2.660 1.0932E-02 1.5179E-01

3.530 1.0671E-03 1.2311E-01

4.080 1.2096E-04 1.0597E-01

4.450 1.3148E-05 9.4996E-02

4.745 1.0253E-06 8.6653E-02

AEmale2 2.660 1.1064E-02 1.5163E-01

3.530 1.0716E-03 1.2297E-01

4.080 1.1415E-04 1.0593E-01

4.450 1.0590E-05 9.6452E-01

4.745 1.0473E-06 8.6232E-02

Study on Speech Transmission under Varying QoS Parameters in a OFDM Communication System

643

Table 5: Results for transmitted speech samples in

British English.

File name SNR

[dB]

BER

after FEC

BER

before FEC

ENfemale1 2.660 1.1316E-02 1.5173E-01

3.530 1.0991E-03 1.2306E-01

4.080 1.1305E-04 1.0589E-01

4.450 1.1924E-05 9.4968E-02

4.745 1.1621E-06 8.6317E-02

ENfemale2 2.660 1.1182E-02 1.5170E-01

3.530 1.0625E-03 1.2297E-01

4.080 1.2698E-04 1.0587E-01

4.450 1.0897E-05 9.5876E-02

4.745 1.0247E-06 8.6664E-02

ENmale1 2.660 1.0953E-02 1.5168E-01

3.530 1.0119E-03 1.2299E-01

4.080 1.1090E-04 1.0591E-01

4.450 1.0980E-05 9.4998E-02

4.745 1.3083E-06 8.5947E-02

ENmale2 2.660 1.1126E-02 1.5177e-01

3.530 1.0424E-03 1.2314E-01

4.080 1.2626E-04 1.0606E-01

4.450 1.1053E-05 9.5845E-02

4.745 1.1258E-06 8.6790E-02

Table 6: Results for transmitted speech samples in

German.

File name SNR

[dB]

BER

after FEC

BER

before FEC

GEfemale1 2.660 1.1190E-02 1.5169E-01

3.530 1.0604E-03 1.2308E-01

4.080 1.0578E-04 1.0602E-01

4.450 1.0543E-05 9.4999E-02

4.745 1.0241E-06 8.6644E-02

GEfemale2 2.660 1.1114E-02 1.5177E-01

3.530 1.0345E-03 1.2307E-01

4.080 1.0335E-04 1.0594E-01

4.450 1.1356E-05 9.6466E-02

4.745 1.2803E-06 8.6679E-02

GEmale1 2.660 1.1301E-02 1.5172E-01

3.530 1.0926E-03 1.2304E-01

4.080 1.1373E-04 1.0592E-01

4.450 1.1790E-05 9.5002E-02

4.745 1.0334E-06 8.6656E-02

GEmale2 2.660 1.1299E-02 1.5181E-01

3.530 1.0610E-03 1.2317E-01

4.080 1.1319E-04 1.0613E-01

4.450 1.0558E-05 9.5014E-02

4.745 1.1809E-06 8.6596E-02

According to obtained results, FEC had an

enormous impact on the final error rate. The initial

range of approx. 10

-1

and 10

-2

was shifted to a new

range from 10

-2

up to even 10

-6

. What is worth

mentioning, this increase was observable regardless

of the audio sample.

The number of erroneous bits has been reduced

by 10 times, in case of the lowest SNR value of

2.660 dB. Whereas, in case of the highest SNR value,

that is 4.745 dB, the quality has been raised more than

a thousand times.

It can be seen that independently from the type of

lector, that is either a female or male individual,

obtained BER values are close in range. In case of

both after and before FEC, obtained results most often

differ only at the second or third decimal position.

Table 7: Results for transmitted speech samples in

Polish.

File name SNR

[dB]

BER

after FEC

BER

before FEC

PLfemale1 2.660 1.0930E-02 1.5182E-01

3.530 1.0603E-03 1.2309E-01

4.080 1.0517E-04 1.0592E-01

4.450 1.1673E-05 9.4971E-02

4.745 1.0515E-06 8.6651E-02

PLfemale2 2.660 1.1021E-02 1.5179E-01

3.530 1.0678E-03 1.2312E-01

4.080 1.1854E-04 1.0560E-01

4.450 1.0614E-05 9.4982E-02

4.745 1.4726E-06 8.6631E-02

PLmale1 2.660 1.1245E-02 1.5172E-01

3.530 1.0859E-03 1.2303E-01

4.080 1.2338E-04 1.0588E-01

4.450 1.1215E-05 9.5551E-02

4.745 1.0268E-06 8.6481E-02

PLmale2 2.660 1.1096E-02 1.5177E-01

3.530 1.0816E-03 1.2309E-01

4.080 1.2623E-04 1.0597E-01

4.450 1.1520E-05 9.5605E-02

4.745 1.1515E-06 8.6650E-02

Not surprisingly, SNR had a profound impact on

the overall error rate. Similar results may be observed

regardless of the language set. An overall summary of

the dependency of BER on SNR is shown in

Figure 2.

QQSS 2021 - Special Session on Quality of Service and Quality of Experience in Systems and Services

644

Figure 2: Dependency of BER on SNR with and without the application of FEC.

5 SUMMARY

As shown, FEC correction coding can raise the

quality of audio content processed and transmitted via

a OFDM communication channel. Depending on the

initial SNR value, this increase may be equal to ten

times or even more than a thousand times.

Furthermore, it has been proved that additional error

correction mechanisms can lower the BER in case of

all speech samples, regardless of the type of lector

(female or male) or even spoken language (American

English, British English, German, and Polish).

As observed, similar results were obtained for all

language sets. A small difference was only noticed on

the second or third decimal point.

Additionally, this small difference in obtained

BER, with and without FEC coding, when comparing

each and every language set with each other, proved

the correctness and accuracy of implementation of the

transmission link. Moreover, it may be said that other

language sets, more popular in other regions of the

world, would perform similarly as the chosen and

investigated portion of samples. This fact makes

future studies, including various additional quality

aspects, even more promising. One should keep in

mind that, all in all, each system and service is

designed to operate and interact with human

end users. This implies many technical aspects,

especially dependability and reliability (Zamojski et

al., 2020).

Certainly, it would be interesting to determine the

impact of SNR and BER, related with objective QoS,

on the subjective judgements of the end user, referred

to as QoE (Quality of Experience). Such an

investigation would surely be interesting to both

researchers and professionals active in the content

creation and distribution link. A good source of

inspiration may be found in (Jeena Jacob,

Kolandapalayam Shanmugam, Piramuthu and

Falkowski-Gilski, 2021). Future studies should and

will therefore focus on evaluating the set of processed

signal samples in a subjective listening quality

evaluation study. It would be interesting to directly

link both SNR and BER parameters with a standard

MOS (Mean Opinion Score) judgement.

It is worth mentioning that the lowest bitrate of

audio content, offered via terrestrial broadcasting as

well as online streaming services, is equal to

32-48 kbps, regardless of the utilized codec. There is

a noticeable trend, that platforms tend to switch from

the well-known MP3 format to AAC (Advanced

Audio Coding), including its numerous profiles,

as well as Ogg Vorbis. Due to the availability and

openness of the Ogg Vorbis format, it seems only a

matter of time when a broader range of platforms will

offer content processed in this particular audio

format. This remark is given not only to commercial

content distribution solutions, but also custom and

dedicated Linux-based ones. Another group of

recipients would be especially those related with the

Study on Speech Transmission under Varying QoS Parameters in a OFDM Communication System

645

IoT (Internet of Things) concept, as well as solutions

operating in the VLSI (Very Large Scale Integration)

and SoC (System on Chip) devices.

REFERENCES

3GPP Technical Specification 23.203. (2011). Technical

specification group services and system aspects;

Policy and charging control architecture.

Agarwal, A., Patra, S. (2012). Performance analysis of

OFDM based DAB systems using concatenated coding

technique. Advanced Materials Research, 403,

105-113.

Aragón-Zavala, A., Angueira, P., Montalban, J., Vargas-

Rosales, C. (2021). Radio propagation in terrestrial

broadcasting television systems: A comprehensive

survey. IEEE Access, 9, 34789-34817.

Boz, E., Finley, B., Oulasvirta, A., Kilkki, K., Manner, J.

(2019). Mobile QoE prediction in the field.

Pervasive and Mobile Computing, 59, 101039.

Cioni, S., Corazza, G. E., Neri M., Vanelli-Coralli, A.

(2016). On the use of OFDM radio interface for satellite

digital multimedia broadcasting systems. International

Journal of Satellite Communications and Networking,

24(2), 153-167.

Clark Jr., G. C., Cain, J. B. (1981). Error-correction coding

for digital communications, Springer. New York,

edition.

ETSI Standard EN 300 401. (2001). Radio broadcasting

systems; digital audio broadcasting (DAB) to mobile,

portable and fixed receivers.

Falkowski-Gilski, P., Uhl, T. (2020). Current trends

in consumption of multimedia content using

online streaming platforms: A user-centric survey.

Computer Science Review, 37, 100268.

Harada, H., Prasad, R. (2002). Simulation and Software

Radio for Mobile Communications, Artech House.

Norwood, 1

edition.

Hossfeld, T., Keimel, C., Hirth, M., Gardlo, B., Habigt, J.,

Diepold, K., Tran-Gia, P. (2014). Best practices for

QoE crowdtesting: QoE assessment with

crowdsourcing. IEEE Transactions on Multimedia,

16(2), 541-558.

ITU Recommendation P.501. (2017). Test signals for

telecommunication systems.

Jeena Jacob, I., Kolandapalayam Shanmugam, S.,

Piramuthu, S., Falkowski-Gilski, P. (Eds.) (2021).

Data Intelligence and Cognitive Informatics:

Proceedings of ICDICI 2020, Springer. Singapore,

edition.

Kim, G., Lee, Y. T., Park, S. R., Lee, Y. H. (2014).

Design and implementation of a novel emergency

wake-up alert system within a conventional T-DMB

service network. IEEE Transactions on Consumer

Electronics, 60(4), 574-579.

Kosaka, A., Okuhata, H., Onoye, T., Shirakawa, I. (2005).

Design of Ogg Vorbis decoder system for embedded

platform. IEICE Transactions on Fundamentals of

Electronics, Communications and Computer Sciences,

88(8), 2124-2130.

Kosaka, A., Yamaguchi, S., Okuhata, H., Onoye, T.,

Shirakawa, I. (2002). VLSI implementation of Ogg

Vorbis decoder for embedded applications.

In: ASIC’02, 15th Annual IEEE International

ASIC/SOC Conference, 20-24.

Kostek, B. (2019). Music information retrieval – the impact

of technology, crowdsourcing, big data, and the cloud

in art. Journal of the Acoustical Society of America,

146(4), 2946-2946.

Zamojski, W., Mazurkiewicz, J., Sugier, J., Walkowiak, T.,

Kacprzyk, J. (Eds.) (2020). Theory and Applications of

Dependable Computer Systems: Proceedings of the

Fifteenth International Conference on Dependability of

Computer Systems DepCoS-RELCOMEX, June 29 –

July 3, 2020, Brunów, Poland, Springer. Cham,

edition.

Zhang, H., Wang, H., Wang, G., Lu, M. (2017). Design and

implementation of the DAB/DMB transmitter

identification information decoder. International

Journal of Circuits, Systems and Signal Processing,

11, 59-64.

QQSS 2021 - Special Session on Quality of Service and Quality of Experience in Systems and Services

646