CC BY
19
Mirsagdiyev Orifjon Alimovich, assistant, of the Department "Electrical connection and radio", Tashkent Institute of Railway Transport Engineering E-mail: [email protected]
EVALUATION OF THE QUALITY OF VOICE TRANSMISSION IN THE NETWORK OF AN OPERATIVE TECHNOLOGICAL NETWORK WITH PACKET SWITCHING
Abstract: Today the networks of operational and technological communications are built using digital channel switching systems. In the future, operative technological network (OTN) networks should implement packet-switching systems. The transition to packet technologies leads to a change in the conditions for voice transmission, so it is necessary to analyze the quality of voice transmission in the prospective OTN networks.
Keywords: operative-technological communication networks, packet switching, voice quality, E-model.
Compared with OTN digital networks with channel switching in packet switched networks, it is possible to distinguish the following features of network construction from the point of view of voice transmission: a central conference server is used; speech transformation can occur in many ways, including coding with accumulation of speech elements with linear prediction; VAD voice activity detectors can be included in conversational tracts.
The quality ofvoice transmission is estimated by the integral indicator - R factor, which is calculated using
the E-model. On the basis of the R-factor, the subjective indicator of the quality of voice transmission is determined - MOS (Mean Opinion Score).
For the calculation of the R-factor and the MOS index, the methodology given in recommendations [1 and 2] and described in [3, 4, 5] is used.
Let's estimate the quality of voice transmission in the circle of the train dispatch communication (TDC). Figure 1 shows a diagram of the organization of a TDC circle in a packet-switched network. It shows only the elements associated with the transmission of speech.
St.1
* £ SDO
OC-E1
Figure 1. Organization of a TDC circle in a packet-switched network
The train dispatch of the circle is in the Single Dispatch Control Center (SDCC), and the duty attendants subordinate to him - at stations 1 ... N, as well as at other stations of direction 2. Each TDC subscriber has one operational console (OC) interworking console, and the dispatcher has a administrative console (OC-A), and the duty officer has an executive console (OC-E) type. The OC is an IP phone that performs the specific functions of the OTN. There are a SDCC conference server (CS) and
a switch (SW), which is used for combining the control console of all dispatchers of the SDCC. The CS server is connected to the station consoles 1 ... N by the IP network, which includes, first of all, routers and switches. The network in question uses Ethernet channels.
In (Figure 2), for the considered TDC circle, a diagram of the colloquial path formed between the OC-A and the OC-E is shown. Voice transmission can occur in both directions in half-duplex mode.
SLRqc 0dB
OC-A
| AM |
g-g &
- g...g
f IP network A
kOi
Q- jA^" JdCJ-0
hs-&sw«i
_SW
tb tam ta
© adder
Figure 2. Scheme of the colloquial tract formed between the OC-A and the OC-E
The structure of the panels and CS includes: a coder (C), a decoder (DC), an adaptation module (AM) and buffers (B). The coder and decoder form a codec that converts speech in the form of a speech wave or a hybrid one in which linear prediction and speech waveform coding are combined. The first codecs include: G.711 and G.721, using pulse-code modulation (PCM) and additive differential PCM (ADPCM), respectively. Hybrid codecs make up the G.72X group, among which G.723.1, G.728 and G.729 codecs are often used.
The adaptation module serves to smooth out the jitter in the speech reception path, while AM accumulates several speech packets. The buffers serve to form the queue of packets when receiving and transmitting. The CS includes digital adders that provide a mode of conferencing and distribution of voice packets in different directions of communication.
Two variants of construction of CS are considered: with adaptive and non-adaptive adders. In the first case, a speech detector is installed at each input of the adder. In it, the summation process occurs when there are speech signals on two or more inputs of the totalizers.
If the signal is only on one input, then the adder sends it to the output without any transformations. A non-adaptive adder does not have a speech detector and summation occurs even if the speech signal is present
only on one input. At the same time, distortion is introduced into the speech signal.
In all the variants of voice transmission, the following indicators are adopted: masking of the local effect -STMR = 15 dB, local effect for the listener - LSTR = 18 dB, D - factor for the subscriber device in transmission and reception - Ds = Dr = 3, threshold noise on the side r eception - Nfor = -64 dB, room noise on the transmitting and receiving sides - Ps = Pr = 35dB, the advantage factor - A = 0.
First, we determine the quality of the transmission with adaptive adder in an individual conversation between the train dispatch communication (TDC) and one station duty officer (SDO), assuming that the speech is transferred from the TDC to the SDO.
Each console is characterized by volume levels: for the transfer SLR= 7dB and for the reception
RLR = 3dB. In the talk channel in the direction from
oc
OC-A to OC-E, the volume level for transmission is:
SLRa = SLR , receive volume level: RLR = RLR . VolA oc B oc
ume levels in the opposite direction will be the same: SLRB = SLRpos = 7dB; RLRA = RLRc= 3dB. The values of the echo reduction level for the speaker TELR and the attenuation of the weighted echo WEPL are determined by the following formulas:
TELR= SLR + RLR + a = 7+ 3 + 75= 85 aE (1)
A A ta " v 7
A
SW
noe-M
Direction 2
ta tb
t.
tb tam
t
ta t
WEPL = 2ata = 150dE, (2)
ata - the transient attenuation between the receiving and transmitting paths inside the OC, the value of ata is assumed equal to 75 dB.
The magnitude ofthe quantization distortion qdu = 1. Taking into account that adaptive adders are used in the conference server, the noise of one direction does not fall into the channel of the other direction of communication. Then the noise level at 0dB: N = -70dB.
c
During the formation, transmission and processing of speech packets in the talk channel, the following delays appear: ta - accumulation delay in the codec (coder or decoder); tb - delay in the receive or transmit buffer; t ,,- delay in the adder; t - delay in AM; t - delay in the
add ' 'am ' 1 n '
IP-network; tc - delay in the OC; ts - delay in the switch.
In the calculations the following ratios of the delay times are accepted: t, = t; t = 2t. The accumulation delay
£ b a' am a '
ta depends on the type ofcodec. The duration tn is a variable, depending on the extent and number of elements ofthe IP network through which the connection is established. In accordance with [2], the time t = 1.5 ms and t ,, = 1 ms.
L J/ oc add
The delay in the switch tk can be assumed to be zero.
Consider the options for voice transmission with codecs G.711 and G.729. In both cases, the accumulation delay ta is assumed to be 10 ms. Variants differ in the hardware distortion index Ie, which is equal to: for codec G.711-0, and for codec G.729-10. Total delay time in the conversation: T=2t + 4t + 4t. + 2t + t,,+1 + t= 3 + 40 + 40 + 40 +
b am add s n
+ 1 + t = 124 + t (mc)
n n v '
The parameter tn includes delays in the nodes of the IP network, and also the propagation time of the signal, depending on the length of the communication channel within the IP network. The value of tn depends to a large extent on how the IP network is constructed, on the number of switching nodes, and on the length of the IP network. In networks with packet technology, the quality of transmission can be significantly affected by the probability of loss of Ppl voice packets. Usually Ppl varies from 1% to 5%.
Let us analyze how the quality ofvoice transmission depends on the time tn and the probability of loss ofvoice packets in the IP network Ppl.
Table 1 shows the calculated values of the R and MOS indicators in the transmission of speech from the TDC to the SDO with adaptive summers in the talk channel, depending on tn and Ppl.
It should be noted that the actual scheme of speech transmission differs from the E-model. This is because the speech signal reflected at the receiving point is transmitted in the opposite direction and enters the input of the adder having the speech detector. Due to the low level of the reflected speech signal, the speech detector will evaluate it as noise and not pass it to the output. The subscriber at the transmission point will not hear the reflected signal. This feature can lead to improved voice quality. However, due to the fact that the TELR and WEPL parameters are large enough, the transmission quality estimate will not practically differ from that (3) given in (Table 1).
Table 1.
t мs n, Ppl, % ^ мs Codec G.711 Codec G.729
R MOS R MOS
1 92.4 4.40 82.4 4.11
15 2 139 92.6 4.40 82.2 4.10
5 91.3 4.37 81.6 4.08
1 91.7 4.38 81.8 4.09
45 2 169 91.5 4.37 81.6 4.08
5 90.9 4.36 81.0 4.06
1 88.5 4.31 78.6 3.97
85 2 209 88.3 4.30 78.4 3.96
5 87.7 4.28 77.8 3.94
1 78.5 3.97 68.5 3.53
165 2 289 78.3 3.96 68.3 3.52
5 77.6 3.93 67.7 3.49
It should be noted that the actual scheme of speech transmission differs from the E-model. This is because the speech signal reflected at the receiving point is transmitted in the opposite direction and enters the input of the adder having the speech detector. Due to the low level of the reflected speech signal, the speech detector will evaluate it as noise and not pass it to the output. The subscriber at the transmission point will not hear the reflected signal. This feature can lead to improved voice quality. However, due to the fact that the TELR and WEPL parameters are large enough, the transmission quality estimate will not practically differ from that given in (Table 1).
Now consider a variant with non-adaptive adder in the collision path of the TDC circle.
In this case, speech conversion takes place in the adder, therefore it is assumed: for codec G.711 - q = 2,
Table
Ie = 0, for codec G.729 - q = 2, Ie = 20. When calculating the noise in the Nc channel, it is necessary to take into account the branch formed in direction 2. The noise increases and amounts to: N = -67dB. The remaining parameters remain the same.
The results of the calculation for the transfer of speech from the TDC to the SDO for the variant with non-adaptive adders are given in (Table 2). Since the MOS value depends little on the probability of packet loss Ppl, calculations are made only for two values of this probability.
The quality of voice transmission in the opposite direction, from TDC to the SDO, in both variants of construction of adders will be the same, which is explained by the inclusion of identical OC consoles on the ends of the conversational paths.
t MS n, Ppl, % ^ мs Codec G.711 Codec G.729
R MOS R MOS
15 1 139 90.7 4.36 70.7 3.63
5 89.8 4.33 70.0 3.60
45 1 169 89.9 4.34 69.3 3.56
5 88.4 4.30 68.6 3.53
85 1 209 86.5 4.24 66.5 3.43
5 85.0 4.20 65.2 3.37
165 1 289 76.1 3.87 56.2 2.90
5 75.3 3.83 55.5 2.87
Conclusion
1. In all cases, the quality of voice transmission is almost independent of packet loss with a loss probability of not more than 5%.
2. The use of G.711 codecs in comparison with G.729 codecs allows to improve the quality of voice transmission, and in the variant with adaptive adder, the quality is improved by (7-8)%, and in the variant with nonadap-tive adders - by (20-60)%.
3. High quality of voice transmission is achieved when using adaptive adders, and it is the same in dispatching circles with different subscriber devices - both consoles and analog telephones. In this case, in accordance with the existing categories of voice quality [3], G.711 codecs achieve a high category, and G.729 codecs are of medium category.
4. When using non-adaptive adders, the transmission quality is reduced. For PDS with G.711 codecs, depend-
ing on the length of the circle, you can achieve an average or high quality category, and with G.729 codecs, the quality will correspond to an unacceptable or a low category. If the subscribers of dispatcher communications have analog telephone sets, the quality is significantly dependent on the direction of voice transmission. When transferred from the controller to the subscriber with G.711 codecs, it is judged to be an unacceptable category. With G.729 codecs, voice quality is unacceptably low. Such a low quality is explained by the small value of the TELR parameter. In the direction of transmission from the subscriber to the dispatcher, the transmission quality becomes higher: with the G.711 codecs, the middle or high category; with codecs G.729 - basically low or middle categories.
5. To achieve voice quality not lower than the middle category (MOS > 3.6), OTN packet networks must be built on the basis of conferencing servers with adaptive
adders. In the conversational tracts of dispatching circles, it is better to use the G.711 codecs, which will almost always give the quality of the highest or highest category. It is important to note that the standard subscriber gateways AG can be used in the OTN network with adaptive
adders, which will significantly affect the cost of OTN systems. OTN networks with nonadaptive adders can be built only with the use of specialized gateways, in the subscriber terminals of which speech detectors should be used.
References:
1. Recommendation ITU-T G.107. The E-model: a computational model for use in transmission planning.- 2009.
2. Recommendation ITU-T G.108. Application of the E-model: A planning guide. - 1999.
3. Лебединский А. К. Оценка качества передачи речи в сетях коммутации каналов и пакетов.- Автоматика, связь, информатика,- 2011.- № 11.- С. 6-9.
4. Лебединский А. К. Мирсагдиев О. А. Оценка качества передачи речи в сети ОТС.- Автоматика, связь, информатика,- 2012.- № 7.- С. 2-5.
5. Рахмангулов А. Н., Мирсагдиев О. А. Имитационная модель оценки качества передачи речи в сетях оперативно технологической связи на железнодорожном транспорте.- Вестник МГТУ им. Г. И. Носова,- 2015.-№ 2.- С. 12-20.
