1 of 22

Analysis of the Cost of Providing User

Perceived Quality in Video on Demand

Presented

by

Telecommunication and Wireless Systems Management

Queen Mary University of London

Ali Adib Arnab | 180656525 | ec18463@qmul.ac.uk

Supervisor: Dr. John Schormans

Second Supervisor: Prof. Pasquale Malacaria

Supervised

by

Edit the text with your own short phrase.

The animation is already done for you; just copy and paste the slide into your existing presentation.

2 of 22

Outline

Introduction

Project Objectives

Literature Review

Related Works

Implementation Setup

Implementation

Discussion and Further Work

References

Experimental Result and Analysis

3 of 22

Introduction

  • Quality of Experience (QoE) has strong correlation to the Quality of services.
  • QoE is forced to be in a complex relationship with
  • QoS factors- packet loss probability (PLP), delay jitter or delay due to network.
  • The evaluation of QoE requires rigorous analysis.
  • Mean Opinion Score (MOS) refers to as an indicator for Quality of Experience (QoE).

​

​

4 of 22

Introduction (Cont…)

  • Analysis of the compound relationship of MOS (of different sorts of TCP sources) with changing Bandwidth cost would be beneficial to evaluate QoE more efficiently.
  • The project gives overview of how within an allocated budget, the amount of MOS score to expect.

​

​

5 of 22

Project Objectives

    • To identify the mathematical relationship between MOS and bandwidth cost by analysing the result.
    • To understand the effect of quality of experience (QoE) and customer satisfaction.
    • Firm knowledge about Mean Opinion Score (MOS) and its indicator and how it is used to measure the quality of experience (QoE)
    • A better understanding of Packet loss and packet loss probability and the relation between packet loss probabilities with Mean Opinion Score (MOS).
    • To obtain changing behaviour of bandwidth with respect to cost for different infrastructures.
    • Getting familiar with MATLAB for arithmetic calculation and curves.

6 of 22

Literature Review

QoE and QoS

  • Quality of Experience (QOE) is now used to evaluate the user-perceived experience (degree of satisfaction/annoyance) of a networked service, e.g. Skype, streaming video, VoIP, etc. [1]

​

  • Quality of Service (QOS) means - "network QOS" are the values of the key network engineering measures throughput, delay, jitter, and packet loss probability [1]

MOS

PLP

Bandwidth Cost

‘Broadband Enhanced & Value’ under BT costs 42£ per month with download speed up to 17 Mbps and upload speed up to 1 Mbps. Another deal called ‘Superfast Fibre Enhanced & Value Line’ costs 57£ per month with download speed up to 76 Mbps and upload speed up to 19 Mbps. It is a measurement of Bandwidth price offered to the customers in the UK. [5]

​

Bandwidth

  • In Digital devices, the bandwidth is generally measured in bits per second(bps) or bytes per second . [4]

​

  • In Analog devices, the bandwidth is measured in cycles per second, commonly known as Hertz. [4]
  • Mean Opinion Score (MOS) refers to as an indicator for Quality of Experience (QOE)
  • Mean Opinion Score within 3-4 means quality is good and 4-5 means excellent. [2]
  • MOS is one way of classifying the characteristic of a phone call. [2]
  • MOS never goes below 1, or above 5. [2]

​

  • Transmission Control Protocol (TCP) is entirely dependent on the average probability (q) of packet loss. [3]
  • There are two parts of Packet loss contingency- packet loss probability in the network and the probability that is experienced by packet at an excessive delay. [3]

7 of 22

Literature Review

Round Trip Time

  • Round-trip time or RTT is also called round-trip delay.
  • It is the time needed in order to a signal pulse or packet to transit from a certain origin to a certain objective and back again. [6]
  • During this context, the origin is that the computer leading off the signal and also the objective could be a remote computer or system that gets the signal and resends it [6]

Bottleneck Capacity in Packets

  • A network bottleneck directs to a unique condition in which computer or network resources restrict the data flow. [7]
  • Bottleneck perpetrator is the network data disruption occurred by microprocessor equipment as of yet TCP/IP. [7]
  • We also have known a network bottleneck as a bottleneck or hot spot. [7]

Packet Buffer Length

  • A packet buffer is memory house put aside for stocking packets awaiting transmission across the networks. [8]
  • The packet delays effects and the streaming for packet loss are being reduced by the Packet buffering in the media systems. [8]
  • In order to synchronize packets, the buffering provides the required time and request and take the place of those lost throughout the transmission [8]

8 of 22

Related Works

  • Roshan and colleagues assessed the relationships between QoS metrics packet loss probability (PLP) and the user-perceived Quality of Experience (QoE) for video on demand (VoD) (Roshan, Schormans and Ogilvie, 2018). [9]
  • In 2008, another work by Xiao focused on TCP connection and network congestion.[10]

Very few accomplished studies evaluate the association between bandwidth cost with QoE.

9 of 22

Implementation Setup

Required Tools and Equations

[9]

MATLAB

    • MATLAB is termed as matrix laboratory. MATLAB is not only a tool for doing math problems.
    • It’s a language which combines visualization, computation and programming in a friendly and easier way to the users
    • MATLAB language is very simple to understand and implement, for some people it is like a calculator which allows user to enable programming as well.
    • MATLAB language is very simple to understand and implement, for some people it is like a calculator which allows user to enable programming as well.
    • Many industries are working based on MATLAB applications. It is used in advanced experimentation and analysis.
    • There are a large number of industries that relies on MATLAB.
    • It is also widely used in academic purposes. For mathematics, engineering and science MATLAB is a standardized tool which is used in schools, colleges and universities.
    • It is also used to conduct scientific research

​

​

​

​

​

​

​

​

​

10 of 22

Implementation

 

11 of 22

Implementation

Screenshots of Codes

​

​

​

​

​

​

​

​

Fig: Determining PLP, MOS value with same Q Fig: Determining PLP, MOS value with different Q

​

​

​

​

​

12 of 22

Applying Curve fitting formula from Cost vs Bandwidth to obtain a formula for MOS vs Bandwidth cost

​

Implementation

13 of 22

Fit name

Data

Fit type

SSE

R-square

DFE

Adj R-sq

RMSE

#coeff

Fit 1

Cost vs BW

Power1

38.4806

0.9589

7

0.9530

2.3446

2

Method

NonlinearleastSquares

Robust

LAR

Algorithm

Trust Region

Implementation: Curve Fitting Tool

14 of 22

Implementation: Goodness of fit

  • f(x)=a*x^b

Coefficients (with 95% confidence bounds)

​

𝐶𝑜𝑠𝑡=27.13∗〖𝐵𝑊〗^0.0986

​

  • SSE: 38.48 (Better SSE if it is closer to 0)
  • R-square 0.9589 (Better R- Square is closer 1)
  • Adjusted R-square 0.953 ((Better R- Square is closer 1))
  • RMSE 2.345

15 of 22

Fig: Bandwidth vs Bandwidth Cost Fig: MOS vs Bandwidth Cost

Implementation

16 of 22

Experimental Result & Output

Fig: MOS vs Bandwidth (Bps) with different packet buffer lengths

Fig: MOS vs Bandwidth (Mbps) with the same packet buffer length

​

17 of 22

Experimental Result & Output

Fig: Bandwidth vs Bandwidth cost relationship

Fig : MOS vs Bandwidth cost relationship

18 of 22

Experimental Result & Output

Fig: MOS vs Bandwidth Cost Output with different TCP sources( N=80)

Fig: MOS vs Bandwidth Cost Output with different TCP sources(N=500)

​

​

19 of 22

Conclusion

Contributions:

  • Found the relationship between MOS and BW cost.
  • Obtained values for graphical illustration.
  • A proportional relationship between MOS and QoE indicating that the better QoE can be obtained from higher Bandwidth. However, more bandwidth leads to more bandwidth cost.

​

20 of 22

Conclusion

Limitations and Future work:

​

  • It is possible to calculate for UDP VOD in the same manner as TCP VOD with few changes.
  • We can work with a bigger range of Bandwidth to determine more practical result.
  • Python and clustering methods can be used to calculate result and it can be compared with MATLAB result to measure accuracy of the project.

​

​

​

21 of 22

References

[1] Zarki, M. (2019). QoS and QoE

[2] Streijl, R., Winkler, S. and Hands, D. (2014). Mean opinion score (MOS) revisited: methods and applications, limitations and alternatives. Multimedia Systems, [online] 22(2), pp.213-227. Available at: https://stefan.winklerbros.net/Publications/mmsj2016.pdf [Accessed 11 Aug. 2019].

[3] Medhi, D. and Ramasamy, K. (2018). Network routing. 2nd ed. San Francisco: Morgan Kaufmann, pp.214-258.

[4] Li, Z., Liu, J. and Drew, M. (2016). Fundamentals of Multimedia. 2nd ed. Vancouver: Springer-Verlag New York Inc, p.152.

[5] British Telecommunications (2019). Business broadband deals | BT Business. [online] Business.bt.com. Available at: https://business.bt.com/products/broadband-and-internet/deals/ [Accessed 14 Aug. 2019.

[6] Rouse, M. (2019). Network management and monitoring: The evolution of network control DEFINITION. [online] SearchNetworking. Available at: https://searchnetworking.techtarget.com/definition/round-trip-time [Accessed 11 Aug. 2019].

[7] Techopedia (2017). What is a Network Bottleneck? - Definition from Techopedia. [online] Techopedia.com. Available at: https://www.techopedia.com/definition/24819/network-bottleneck [Accessed 9 Aug. 2019].

[8] Techopedia (2017). What is Packet Buffer? - Definition from Techopedia. [online] Techopedia.com. Available at: https://www.techopedia.com/definition/2796/packet-buffer [Accessed 11 Aug. 2019].

[9] Roshan, M., Schormans, J. and Ogilvie, R. (2018). Video-on-demand QoE Evaluation Across Different Age- Groups and Its Significance for Network Capacity. ICST Transactions on Mobile Communications and Applications, [online] 3(9), p.153557. Available at: https://www.researchgate.net/publication/322388390_Video-on-demand_QoE_Evaluation_Across_Different_Age-_Groups_and_Its_Significance_for_Network_Capacity [Accessed 9 Aug. 2019].

[10] Xiao, X. (2008). Technical, commercial and regulatory challenges of QoS. Amsterdam [u.a.]: Elsevier/Morgan Kaufmann, p.30.

​

​

​

​

​

​

22 of 22

THANK YOU