VoIP Şebekesindeki Ses Kalitesinin Çok Katmanlı Yapay Sinir Ağları ile Modellenmesi

Öz

Son yıllarda internet ağının, çok yaygınlaşması, esnek olması, hızlı ve yüksek kapasite erişimi sunabilmesine bağlı olarak farklı tipteki birçok servis internet protokolü (IP) üzerinden iletilmeye başlanmıştır. Bu servis tiplerinden birisi de ses trafiği olup bu durum klasik haberleşme sistemlerini doğrudan etkilemektedir. Ses iletimi, genel anahtarlamalı telefon ağları (public switched telephone network, PSTN) yerine IP üzerinden yapılmaya başlanmıştır. Bu yeni haberleşme teknolojisi, IP üzerinden ses iletimi (voice over IP, VoIP) olarak adlandırılmıştır. VoIP teknolojisinde, ses iletiminin kullanıcı tarafından algılanma kalitesi ise ortalama görüş puanı (mean opinion score, MOS) ile belirlenir. MOS değerinin üretilmesi birçok parametreye bağlıdır. ITU-T (international telecommunications union -telecommunication standardization sector) tarafından MOS değerinin belirlenmesi için PSQM (perceptual speech quality measure), PESQ (perceptual evaluation of speech quality), POLQA (perceptual objective listening quality analysis) gibi birçok model standartlaştırılmıştır. Fakat bu model yapıları, birçok avantaja sahip olmasının yanında referans işaret gereksiniminden dolayı uygulamalarda pratik zorluklar oluşturmaktadır. Buyüzden bu çalışmada, ses iletimindeki MOS kalite değerlerini tahmin etmek amacıyla referans sinyal gerektirmeyen çok katmanlı yapay sinir ağları (YSA) tabanlı bir model yapısı tasarlanmıştır. Bu modelde, giriş olarak IP trafiğine ait servis kalite (quality of service, QoS) parametreleri ve çıkış olarak ise elde edilen MOS değerleri kullanılmıştır. Giriş olarak kullanılan bu QoS parametreleri ise paket kaybı (packet loss) ve gecikme (delay) değerleridir. Böylece 2 giriş ve 1 çıkışa sahip bir model oluşturulmuştur. Ayrıca model, farklı QoS parametresi kullanılarak MOS tahmini yapabildiği için esnek bir yapıya da sahiptir. Literatürdeki çalışmalardan farklı olarak bu çalışmada, POLQA yöntemi ile ölçülen MOS değerlerinin YSA model ile tahmin edilmesine çalışılmıştır. YSA modelin eğitim ve test sürecinde LMB (levenberg-marquardt backpropagation), BRB (bayesian regulation backpropagation) ve RSB (Resilient backpropagation) algoritması kullanılmıştır. Simülasyon sonucunda elde edilen sonuçlar tablolar ve şekiller vasıtasıyla sunulmuştur. Sonuçlara göre geliştirilen bu metodun ITU-T tarafından önerilen modeller ile karşılaştırılabilir seviyede olduğu gösterilmiştir.

Anahtar Kelimeler

• VoIP
• Ortalama Görüş Puanı
• POLQA Model
• Yapay Sinir Ağları

Modeling of Sound Quality in VoIP Network with Multi-Layer Artificial Neural Networks

Abstract

In recent years, different types of services have been started to be transmitted over the internet protocol (IP), due to the widespread use of the internet network, its flexibility and the ability to offer fast and high capacity access. One of these service types is voice traffic and this directly affects classical communication systems. Voice transmission is started over IP instead of public switched telephone network (PSTN). This new communication technology is called voice over IP (VoIP). In VoIP technology, the quality of the perception of the voice transmission by the user is determined by the mean opinion score (MOS). Generation of the MOS value depends on many parameters. Many models such as PSQM (Perceptual Speech Quality Measure), PESQ (Perceptual Evaluation of Speech Quality), POLQA (Perceptual Objective Listening Quality Analysis) have been standardized by ITU-T (International Telecommunications Union -Telecommunication Standardization Sector) to determine the MOS value. However, besides having many advantages these model structures also create practical difficulties in applications due to the need of a reference signal. Therefore, in this study, a multi-layered artificial neural networks (ANN) based model structure that does not need a reference signal is designed to estimate MOS quality values in voice transmission. In this model, quality of service (QoS) parameters of IP traffic are used as input and obtained MOS values are used as output. These QoS parameters used as input are packet loss and delay values. Thus, a model with 2 inputs and 1 output was created. In addition, the model has a flexible structure as it can estimate MOS using different QoS parameters. Different from the studies in the literature, in this study, it was tried to estimate MOS values measured by the POLQA method with ANN model. LM (Levenberg Marquardt), BR (Bayesian Regulation) and RPROP (Resilient Backpropagation) algorithm was used in the training and testing process of the ANN model. The results obtained from the simulations are presented by tables and figures. The method, developed according to the results, has been shown to be comparable to the models recommended by ITU-T.

Keywords

• Artificial Neural Networks
• POLQA Model
• Mean Opinion Score
• VoIP

Kaynakça

1. Çalık, O., Irıcıoğlu, U., Karabulut Kurt, G., Pusane, A.E., Demiroğlu, A.S., & Kayık, G. (2016). Impact of retransmissions on the quality of experience in VoIP systems. Elektrik-Elektronik-Bilgisayar Mühendisliği Sempozyumu ve Fuarı (ELECO), Bursa, 617-621.
2. Pala, Z. (2017). Kampüs ağlarında arkaplan trafiğin IP-tabanlı telefon sistemlerin ses kalitesi üzerindeki etkisi. Iğdır Üniversitesi Fen Bilimleri Enstitüsü Dergisi, 7(2), 55-63.
3. Al-Wahshat, H., Al-Maitah, M., & Al-Smadi, T. (2017). Voice quality for internet protocol based on neural network model. Journal of Signal and Information Processing, 8, 195-202. (https://doi.org/10.4236/jsip.2017.84013)
4. Agrisani, L., Capriglinoe, D., Ferrigno, L., & Miele, G. (2016). Measurement of the IP packet delay variation for a reliable estimation of the mean opinion score in VoIP services. IEEE International Instrumentation and Measurement Technology Conference Proceedings, Taipei, Taiwan, 1-6. (https://doi.org/10.1109/I2MTC.2016.7520492)
5. Hartpence, B. (2013). Packet Guide to Voice over IP. O’Reilly press.
6. Kadıoğlu, R., Dalveren, Y., & Kara, A. (2015). Quality of service assessment: a case study on performance benchmarking of cellular network operators in Turkey. Turk J Elec Eng & Comp Sci, 23, 548-559. (https://doi.org/10.3906/elk-1302-191)
7. Nipp, O., Kuhn, M., Wittneben, A., & Schweinhuber, T. (2007). Speech quality evaluation and benchmarking in cellular mobile networks. IEEE 2007 Mobile and Wireless Communications Summit, Budapest, Hungary, 1-5. (https://doi.org/ 10.1109/ISTMWC.2007.4299219)
8. Mossavat, I. (2012). A hierarchical Bayesian approach to modeling heterogeneity in speech quality assessment. IEEE T Audio Speech, 20, 136-146. (https://doi.org/ 10.1109/TASL.2011.2158421)

9. ITU-T (1996). Recommendation P. 861, Objective quality measurement of telephone band (300 - 3400 Hz) speech codecs. (https://www.itu.int/rec/T-REC-P.861-199608-S/en), (Erişim Tarihi: Mayıs 2020).
10. Kuipers, F., Kooij, R., De Vleeschauwer, D., & Brunnström, K. (2010).Techniques for measuring quality of experience. International Conference on Wired/Wireless Internet Communications (WWIC), Lulea, Sweden, 216-227.
11. Opticom GmbH. (2005). PESQ - Percepual evaluation of speech quality. (http://www.opticom.de/technology/pesq.php), (Erişim Tarihi: Mayıs 2020).
12. Jelassi, S., Rubino, G., Melvin, H., Youssef, H., & Pujolle, G. (2012). Quality of experience of VoIP service: a survey of assessment approaches and open issues. IEE Communications Surveys & Tutorials, 14, 491-513. (https://doi.org/ 10.1109/SURV.2011.120811.00063)
13. Pocta, P., Cinar, Y., & Melvin, H. (2016). Black-box analysis of the extent of time-scale modification introduced by WebRTC adaptive jitter buffer and its impact on listening speech quality. Journal Communications, 18(1), 17-22.
14. Hines, A., Skoglund, J., Kokaram, A., & Harte, N. (2013). Robustness of speech quality metrics to background noise and network degradations: comparing VISQOL, PESQ and POLQA. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Canada, 3697-3701. (https://doi.org/10.1109/ICASSP.2013.6638348)
15. Gaoxiong, Y., & Wei, Z. (2012). The Perceptual objective listening quality assessment algorithm in telecommunication: introduction of ITU-T new metrics POLQA. 1st IEEE International Conference on Communications in China (ICCC), Beijing, China, 351-355. (https://doi.org/ 10.1109/ICCChina.2012.6356906)
16. POMY, J. (2011). POLQA- The next-generation mobile voice quality testing standard. ZNIIS / ITU Workshop. (https://www.itu.int/ITU-D/tech/events/2011/Moscow_ZNIIS_April11/Presentations/09-Pomy-POLQA.pdf), (Erişim Tarihi: Mayıs 2020).
17. Gerlach, O. (2012). Next-generation (3G/4G) voice quality testing with POLQA. (https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_application/application_notes/1ma202/1MA202_1e_3G4G_voice_quality_testing_POLQA.pdf), (Erişim Tarihi: Mayıs 2020).
18. Mishra, K.C., & Das, P.C. (2015). Measuring quality of service of VoIP based on artificial neural network approach. International Journal of Advanced Research in Computer Science and Software Engineering, 5(3), 657-661.
19. Ren, J., Mao, D., & Wang, Z.W. (2009). A neural network based model for VoIP speech quality prediction. Proceedings of the 2nd International Conference on Interaction Sciences: Information Technology, Culture and Human, Seoul, Korea, 1244-1248. (https://doi.org/10.1145/1655925.1656152)
20. AL-Akhras, M., ALMomani, I., & Sleit, A. (2011). An improved E-model using artificial neural network VoIP quality predictor. Neural Network World, 21(1), 3-26. (https://doi.org/10.14311/NNW.2011.21.001)
21. Sun, L., & Ifeachor, E.C. (2002). Perceived speech quality prediction for voice over IP-based networks. IEEE International Conference on Communications (ICC), New York, USA, 1-5. (https://doi.org/10.1109/ICC.2002.997307)
22. Larijani, H., & Radhakrishnan, K. (2010). Voice quality in VoIP networks based on random neural networks. Ninth International Conference on Networks, Menuires, France, 89-92. (https://doi.org/10.1109/ICN.2010.23)
23. Konar, M. (2019). GAO algoritma tabanlı YSA modeliyle İHA motorunun performansının ve uçuş süresinin maksimizasyonu. European Journal of Science and Technology, 15, 360-367. (https://doi.org/10.31590/ejosat.529093)
24. Haykin, S. (1999). Neural networks-a comprehensive foundation, Prentice Hall, 2nd ed.
25. Öztürk, C. (2011). Yapay sinir ağlarının yapay arı kolonisi algoritması ile eğitilmesi. Erciyes Üniversitesi, Fen Bilimleri Enstitüsü, (Doktora Tezi).
26. Kılıç, E., Özbalcı, U., & Özçalık, H.R. (2012). Lineer olmayan dinamik sistemlerin yapay sinir ağları ile modellenmesinde MLP ve RBF yapılarının karşılaştırılması. Elektrik - Elektronik ve Bilgisayar Mühendisliği Sempozyumu (ELECO), Bursa, 694-698.
27. Bağiş, A., & Konar, M. (2010). Uçuş kontrol sistemi yakıt parametresinin yapay sinir ağları kullanılarak belirlenmesi. Akıllı Sistemlerde Yenilikler ve Uygulamaları Sempozyumu (ASYU’2010), Kayseri, Türkiye, 104-108.
28. Konar, M., & Bagiş, A. (2016). Simultaneous computation of the speed and fuel parameters of flight control system by using Anfis and artificial neural networks. 24th Signal Processing and Communication Application Conference (SIU 2016), 1389-1392. (https://doi.org/10.1109/SIU.2016.7496008)
29. Oktay, T., Arik, S., Turkmen, I., Uzun, M., & Celik, H. (2018). Neural network based redesign of morphing UAV for simultaneous improvement of roll stability and maximum lift/drag ratio. Aircraft Engineering and Aerospace Technology, 90(8), 1203-1212. (https://doi.org/10.1108/AEAT-06-2017-0157)