CN109873797A

CN109873797A - Conversational video business QoE-QoS parameter mapping method based on statistical analysis

Info

Publication number: CN109873797A
Application number: CN201810151474.4A
Authority: CN
Inventors: 董育宁; 曹婷婷
Original assignee: Nanjing Post and Telecommunication University
Current assignee: Nanjing Post and Telecommunication University; Nanjing University of Posts and Telecommunications
Priority date: 2018-02-14
Filing date: 2018-02-14
Publication date: 2019-06-11

Abstract

The invention belongs to video traffic QoE and network resource management field, specifically, it is a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis, first, video data stream in network needed for data acquisition analyzes software grabs by network package, the original data stream of crawl is saved into received text format；Secondly, carrying out a large amount of traffic statistics feature calculation；Then, MOS value is calculated according to the standard meter of International Telecommunication Union；It chooses the parameters such as packet loss, transmission rate and delay variation and carries out graphic plotting and formula fitting, optimal parameter recommendation index is obtained according to result；Finally by give reference value to substitute into model to calculate, statistics falls in the probability of previously given MOS value range and is verified.The present invention can be under the premise of guaranteeing user experience quality, more reasonable effective distribution Internet resources, avoid too much distributing to Internet resources the user that user experience quality has been met so as to cause the wasting of resources.

Description

Conversational video business QoE-QoS parameter mapping method based on statistical analysis

Technical field

The invention belongs to QoE and Internet resources to distribute field, and in particular to a kind of conversational video industry based on statistical analysis Business QoE-QoS parameter mapping method.

Background technique

With the rapid development of computer network and the communication technology, network multimedia business plays in people's lives More important role permeates every aspect in daily life.The type of multimedia service application is very various, such as regards Frequently, medical treatment, shopping online, monitoring, teaching etc..Wherein, with shapes such as video conversation, network direct broadcasting, 2D and 3D big games The session stream development of formula is especially rapid, the following development along with economy and internet, and the market share specific gravity of video conversation will Sustainable growth.It is more rationally efficient to distribute Internet resources how under the premise of guaranteeing user experience quality, it is ISP (network Service provider) the problem of paying close attention to and unusual urgent problems.

In multimedia service, service quality (QoS) may be the most service models used in the past, international telecommunication Alliance (ITU) defines service quality (QoS) in " the open processing reference model of information technology " in x.902 standard are as follows: fixed The set of a set of quality requirement of the justice in the collective behavior of one or more objects, it is emphasised that technical aspect objective finger Mark.As handling capacity, transmission delay, shake, packet loss, bandwidth etc., these QoS parameters just represent the evaluation index of QoS, but It is that requirement with people to transmission services is higher and higher, relies solely on subjectivity of the individual QoS without combining user itself Impression, can not accurately meet the growing Quality of experience requirement of people.And Quality of experience (QoE), refer to user couple The direct feel of network, system or quality of service and performance, it combines the factor of every aspect, therefore can be very good to retouch State the quality of network service.MOS is to indicate that user is the flat of consumers' opinions point to the perceptual parameters of real-time multimedia traffic quality Mean value, value range are 1~5.That is MOS value subjective assessment from the user, its result is the most accurate, but this A large amount of manpower and material resources can be lost in kind method, and real-time is not high.Probability density function is the output valve for describing stochastic variable, The probability function of value near the point that some is determined.

Resource allocation is a very complicated topic, it is related to the every aspect of network.It is primarily referred to as solving to use The problem of family network performance declines, according to the network actual conditions of user, is analyzed, orientation problem mentions from different angles For the suggestion of resource optimization.Since designing first generation network, just there are many being distributed about Internet resources, up to now It is also a popular field.

Currently, the research about QoE is main still in the foundation of model, carried out in other words by already existing model Verifying, rarely probes into the distribution relation of parameters in the case where known MOS value, and it is applied to and meets user experience While quality, Internet resources are reasonably distributed.QoE model itself is more complicated, is a nonlinear optimization problem, mesh Scalar functions are non-convex, ask the process of optimal solution or its inverse function very troublesome.In addition the mapping from parameters to QoE is natively It is more to one mapping, a variable cannot determine QoE completely.So if the distribution from QoE to derive parameter, very Seldom to a correct inverse function or optimal solution.Based on such a case, the present invention just popular now session Class video calling carries out QoE research, and analysis session business QoE model probes into the parameters relationship in each MOS value piecewise interval, And in terms of it being applied to the distribution of Internet resources.The present invention proposes a kind of conjunction on the basis based on user experience quality A kind of method of the effective distribution Internet resources of reason can have been expired to avoid by the excessive user experience quality of distributing to of resource Foot user so as to cause the wasting of resources.

Summary of the invention

Present invention aims at aiming at the problem that above-mentioned multimedia present Research, just conversation class video popular now Call carries out QoE research, and analysis session business QoE model is probed into the parameters relationship in each MOS value piecewise interval, proposed A kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis.Experimental data set has chosen QQ and Skype Video calling, the experimental results showed that parameter distribution has certain rule within the scope of the MOS value in each section.Utilize the present invention Achievement building for different section MOS QoS parameter (packet loss, bandwidth) can be provided under conditions of guaranteeing user experience quality Index is discussed, is suitable for Internet resources distribution aspect, avoiding distributing to Internet resources into user experience quality too much has been expired Foot user so as to cause the wasting of resources.

To achieve the above object, the technical solution adopted by the present invention is that a kind of conversational video business based on statistical analysis QoE-QoS parameter mapping method, includes the following steps:

Step 1: dataset acquisition:

1.1) using the data flow at session service both ends in network needed for network package analysis software grabs, by the original of crawl Beginning data flow is saved into received text format；

Step 2: feature extraction:

2.1) every stream is subjected to fragment by the data flow after crawl；

2.2) special parameter of each flow is counted；

Step 3:MOS value calculates:

3.1) according to ITU-T P1201.1 model and the corresponding MOS value of Parameter Calculation；

Step 4: drafting and analyzed pattern:

4.1) choosing certain parameters is input, three-dimensional, the two-dimensional distribution of MOS value is obtained, to conversational video business datum It is for statistical analysis, it is given at the suggestion reference value of different MOS interval parameters；

4.2) MOS value is segmented, draws the probability distribution histogram of each segmentation MOS interval parameter, and carry out formula Fitting, the indices of digital simulation formula；

4.3) distribution situation of the delay variation in each segmentation section MOS is counted；

Step 5: conclusion and verifying:

5.1) formula for comprehensively considering the figure that step 4.1 and 4.2 etc. are drawn and fitting is given at the different areas MOS Between qos parameter (packet loss, bandwidth) suggestion index.

5.2) the suggestion index provided in step 5.1 is substituted into ITU-T P.1201.1 model, counts each piecewise interval Reach the probability of given MOS value range, the accuracy of index is suggested in verifying.

Further, above-mentioned network package analysis software is WireShark.

Preferably, above-mentioned standard text formatting include packet arrival time, source IP address, purpose IP address, protocol type, Five column raw information of data package size.

Fragment is carried out to data stream described in above-mentioned steps 2.1, is because used conversation class QoE model is suitable in short-term It flows (10s-16s), experiment carries out fragment with 15s.

The special parameter in above-mentioned steps 2.2 includes packet loss, transmission rate, delay variation, interruption etc..

Preferably, ITU-T P1201.1 model described in above-mentioned steps 3.1 is International Telecommunication Union's formulation for network The calculation method of the user experience quality of Streaming Media mainly applies to the media payload being encapsulated in RTP/UDP/IP packet (audio, video) is a kind of regional model algorithm of low resolution (such as mobile TV).

Certain parameters described in above-mentioned steps 4.1 are packet loss and transmission rate.Wherein, choosing packet loss is abscissa, is passed Defeated rate is ordinate.

MOS value is segmented for interval with 1 preferably, MOS value described in above-mentioned steps 4.2 is segmented into.

Preferably, the fitting of formula described in above-mentioned steps 4.2 refers to the form for being fitted to second order Gauss, indices packet It includes: and variance, root mean square and determining coefficient.

The distribution situation of the delay variation in each segmentation section MOS of statistics described in above-mentioned steps 4.3 refers in step 4.2 On the basis of, the distribution situation of delay variation is counted in the segmentation section MOS, mean value and standard deviation is calculated, with the shape of histogram Formula graphing.

Preferably, the suggestion index of qos parameter described in above-mentioned steps 5.1 (packet loss, bandwidth) be according to 4.1 and 4.2 comprehensive one obtained consider the suggestion index of Internet resources and user experience quality.The suggestion index that the present invention is given All be with center at principle obtained by because the index obtained in this way, not only ensure that user experience quality with maximum probability, And also accomplish not waste as far as possible in terms of resource allocation, it avoids distributing to Internet resources into user experience quality too much and has obtained To satisfaction user so as to cause the wasting of resources.4.1 figures drawn can be divided into bandwidth sensitive area and immunity region, have Apparent distribution characteristics.The first step first makees a line of demarcation across each MOS value, and cut-off rule top half is bandwidth sensitive Area, lower half portion are bandwidth immunity region；Second step finds out the midpoint on left side transmission rate boundary, and therefrom point is drawn in one Line mutually gives cut-off rule；Third step finds out the midpoint of the packet loss on the right, and therefrom point draws middle line intersection cut-off rule and same A bit, as parameter recommendation index.Other MOS values section repeats second step, third step.Other business scenarios, repeat more than Step to get each section out parameter reference values.In conjunction with 4.2 PDF function, building for the qos parameter in the different sections MOS is obtained Discuss index.

The accuracy of index is suggested in verifying described in above-mentioned steps 5.2, the specific steps are that will first calculate resulting each group number It is interval with 0.5 according to classifying, is classified according to MOS value；Then the suggestion index in each section MOS is replaced respectively The packet loss and bandwidth of initial data, calculate MOS value；MOS value is counted in the probability in the given section MOS.

Compared with prior art, beneficial effects of the present invention:

1, the present invention can calculate every according to the existing ITU-T Media Stream QoE standard that P.1201.1 user perceives The MOS value of video flowing, the distribution situation in each section of analysis session business MOS value, can not only guarantee client video calling Quality of experience, and the suggestion index of parameter can be provided in the case where each given MOS value, be conducive to the conjunction of Internet resources Reason distribution and utilize, avoid too much distributing to Internet resources the user that user experience quality has been met so as to cause The wasting of resources.

2, the parameter distribution situation in each section MOS of present invention research session service, from having reached on the other hand The problem of how Internet resources under conditions of user experience quality should distribute met, has jumped out the difficult point of the derivation of equation, also more The defect of current this respect technical research has been mended, no matter all there is certain meaning from thought or inventive result.

Detailed description of the invention

Fig. 1 is the flow diagram of session service network parameter predictive model algorithm of the present invention.

Fig. 2 is that data acquisition network of the present invention opens up topological diagram.

Fig. 3 is ITU-T of the present invention P.1201.1 QQ transmission rate, the relational graph of packet loss and MOS value under standard.

Fig. 4 is the PDF analysis chart of QQ parameter distribution of the present invention.

Specific embodiment

The present invention is described in further detail with reference to the accompanying drawings of the specification.

As shown in Figure 1, the invention proposes a kind of conversational video business QoE-QoS parameter mapping side based on statistical analysis Method, this method specifically include data acquisition, feature extraction, parameter and the calculating of MOS value, drafting and analysis of two-dimensional and three-dimensional Figure, the classification of MOS value and Function Fitting, verifying.Main specific steps are as follows:

Step 1, data acquire, and: Wireshark is a kind of network package analysis software, intercept network package, and to the greatest extent may be used Can show the most detailed network package data, it be the widest network package analysis software that uses in the world at present it One.Experimental data of the invention is taken by Wireshark, and the place of experiment is in the comprehensive scientific research of Nanjing Univ. of Posts and Telecommunications Building and students' dormitory, time are in December, -2017 in November, 2017, and what data service was chosen is QQ and Skype video calling, Including wired and wireless data, while being collected in two different subnet computer terminals, network topological diagram such as Fig. 2.

QQ and Skype call is carried out respectively in six and complex building, grabs media session class industry simultaneously using wireshrak Five group informations such as acquisition time, source IP address, purpose IP address, protocol type and the packet size at business stream both ends.Wherein QQ Video 100*2 item stream, Skype88*2 item stream, every stream 30min, details are as shown in table 1:

1 data set of table

Step 2, feature extraction: because ITU-T P.1201.1 standard be applicable in be in short-term stream (10s-16s), need Every stream is subjected to fragment, every 15s is a flow, and the QQ video sample number obtained in this way is 12000*2, Skype sample Number is 10560*2.The parameter at both ends is counted respectively, and finally weighted average obtains packet loss, delay variation and transmission speed The parameters such as rate.Finally obtained number of samples is QQ video 12000, Skype10560.

Step 3, parameter and MOS value calculate: according to ITU-T P1201.1 model, calculating basic parameter, then basis Parameter calculates corresponding MOS value.

Step 4, it draws and analysis of two-dimensional and three-dimensional figure: being input parameter with packet loss and transmission rate, obtain MOS The three-dimensional of value, two dimension PDF distribution map, as shown in figure 3, analyzed pattern and being given within the scope of fixed MOS value, the reference value of parameter (or minimum value and maximum value), to be applied to the reasonable distribution for reaching resource in practice.

In experiment, the transmission mode of wired and wireless situation has been distinguished in all data analyses, has chosen QQ and Skype The mode of video communication is that representative is analyzed respectively.The specific X-Y scheme using Fig. 3 obtains the drawing method of optimal reference point Are as follows: 1) first across each MOS value make a line of demarcation, cut-off rule top half be bandwidth sensitive area, lower half portion be bandwidth not Sensitizing range；2) middle line is drawn from the midpoint of left side transmission rate mutually give cut-off rule；3) from the midpoint of the packet loss on the right Draw a middle line intersection cut-off rule and same point.Other MOS values section repeats second step, third step, obtains each section Parameter reference values.

It can be seen that, wired packet loss, transmission rate, MOS value etc. are all more ideal than wireless data from experimental result, MOS value rises overally than wireless.With Fig. 3 QQ data instance, comparison is wireless and cable data, the packet loss model of QQ wireless data It encloses for 0-10% (most of data concentrate on 0-6%), transmission rate range is that (most of data concentrate on 0-1000kbps 100-500kbps), MOS value range is 1-4 (being largely focused on 2.5-3.5)；And cable data is compared, packet loss range is 0-7% (most of data concentrate on 0-4%), transmission rate range are that 0-1600kbps (is largely focused on 600- 1200kbps), MOS value range is 1-4.5 (being largely focused on 3.5-4.5).It can be seen that cable data is integrally used than wireless data Family experience will be got well.

Step 5, the classification of MOS value and Function Fitting: as shown in figure 4, being respectively divided into 4 using MOS for 1 as section The distribution situation of packet loss under the MOS value in each section, transmission rate and delay variation is studied in a section.According to packet loss Rate, the distribution situation of transmission rate are fitted corresponding formula, and what the form of formula was taken is the form of second order Gauss, letter in experiment The criterion that number is taken has: and variance, root mean square and determining coefficient.Delay variation is counted in the distribution in each section MOS Situation draws histogram, obtains the requirement that delay variation should reach in each section MOS.

Step 6, as a result with verifying: under the premise of meeting user experience quality point, can be given at respectively in conjunction with step 4 and 5 The optimal reference point of a MOS value section difference transmission mode, different sessions business, as shown in table 2.The reference value provided is substituted into Model is verified in each section MOS, counts the probability that each section MOS reaches given MOS value, as shown in table 6.

The suggestion index of the 2 difference section MOS qos parameter (packet loss, bandwidth) of table

Abscissa is packet loss (%), and ordinate is transmission rate (100kbps).

P.1201.1 stream media quality prediction model is described in detail below ITU-T:

P.1201.1, ITU-T is a kind of no reference model, and the media for mainly applying to be encapsulated in RTP/UDP/IP packet have It imitates load (audio, video), is a kind of regional model algorithm of low resolution (such as mobile TV).It includes three modules, point Not Wei audio quality evaluation module, video quality evaluation module and audio-visual quality evaluation module, entire mode input be packet The information and encoding and decoding of head, export as the MOS value of the MOS value of audio, the MOS value of video and audio-video.

Since choose is session service to the present invention, according in ITU-T P.1201.1 streaming media service quality evaluation model Audio-visual quality assessment models, calculate MOS value, then analyze the parameters such as packet loss, transmission rate and delay variation in spy Determine the distribution situation within the scope of MOS value, draws two dimension, three-dimensional, PDF and column distribution map, finally provide in each segment The reference value of parameter.A specific descriptions mainly are made to audio-video assessment models below.

Audio-video MOS value calculates as follows:

AV_MOSC indicates the audio-visual quality of consideration compressive damage, and calculation is as follows:

AV_MOSC=av1*V_MOSC+av2*A_MOSC+av3*V_MOSC*A_MOSC+av4

3 AV_MOSC audio-visual quality metewand collection of table

	QCIF	QVGA	HVGA
				av1	0.7977	0.7495	0.6419
av2	0.03732	0.09736	0.1362
				av3	0.02472	0.006725	0.016
av4	0.1657	0.3186	0.5694

AV_MOSP indicates the audio-visual quality of consideration packet loss damage, and calculation is as follows:

AV_MOSP=AV_MOSC-AV_DP

AV_DP is audio-visual quality caused by considering packet loss:

AV_DP=(AV_MOSC-MOS_MIN) * AV_DF

AV_DF represents distortion factor caused by audio-video packet loss, and calculation is as follows:

AV_DFV represents distortion factor caused by video packet loss, similarly, AV_DFA represent be distorted caused by audio packet loss because Son.

4 AV_DP audio-visual quality metewand collection of table

AV_MOSR indicates the audio-visual quality of consideration packet loss damage, and calculation is as follows:

NRE indicates interruption times, and ARL represents average buffer time, and MREEF represents mean down time.

5 AV_DR audio-visual quality metewand collection of table

Other parameters can be obtained from audio or Video Model.

Experimental result of the invention:

In order to further verify the validity for the statistical nature that the present invention chooses, we use above two session service: QQ and Skype, respectively analyzes cable data and wireless data.Utilize the meeting based on statistical analysis proposed in the present invention Video traffic QoE-QoS parameter mapping method is talked about, achievement of the invention is substituted into model and is calculated, statistics reaches given MOS value model The probability enclosed compares verifying, as shown in chart 6.It can be seen that, by substituting into reference value, reach given MOS value from table The probability of range is very high, it is seen that effect of the invention be it is extraordinary, can be effectively avoided and allocate resources to body The wasting of resources caused by the user that the amount of checking the quality has been met.

The verifying of the 6 difference section MOS qos parameter (packet loss, bandwidth) of table suggestion index

It should be noted that data described above are only obtained by a specific embodiment of the invention, not to limit The present invention, all within the spirits and principles of the present invention, any modification, equivalent replacement, improvement and so on should be included in this Within the protection scope of invention.

Claims

1. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis, which is characterized in that the method packet Include following steps:

Step 1: data acquisition:

1.1) data flow at session service both ends in network needed for network package analysis software grabs, including wired and nothing are utilized Line saves the original data stream of crawl at received text format；

Step 2: feature extraction:

2.1) every stream is subjected to fragment by the data flow after crawl；

2.2) count the special parameter of each flow, the special parameter include packet loss, transmission rate, delay variation, in It is disconnected；

Step 3:MOS value calculates:

Step 4: drafting and analyzed pattern:

4.1) choosing certain parameters is input, obtains three-dimensional, the two-dimensional distribution of MOS value, is carried out to conversational video business datum Statistical analysis, is given at the suggestion reference value of different MOS interval parameters, and the parameter is packet loss and transmission rate, wherein choosing Taking packet loss is abscissa, and transmission rate is ordinate；

4.2) MOS value is segmented, draw it is each segmentation MOS interval parameter probability distribution histogram, and to and to PDF function into Row formula fitting, the indices of digital simulation formula；

Step 5: conclusion and verifying:

5.1) formula for comprehensively considering the figure that step 4.1 and 4.2 etc. are drawn and fitting is given at the different section MOS QoS The suggestion index of parameter；

5.2) the suggestion index provided in step 5.1 is substituted into ITU-T P.1201.1 model, counts each piecewise interval and reaches The accuracy of index is suggested in the probability of given MOS value range, verifying.

2. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, Be characterized in that: the network package analysis software is WireShark.

3. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, Be characterized in that: the received text format is big including packet arrival time, source IP address, purpose IP address, protocol type, data packet Small five column raw information.

4. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, It is characterized in that: fragment being carried out to data stream described in step 2.1, experiment carries out fragment with 15s.

5. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, Be characterized in that: ITU-T P1201.1 model described in step 3.1 is the user for network flow-medium that International Telecommunication Union formulates The calculation method of Quality of experience mainly applies to the media payload being encapsulated in RTP/UDP/IP packet, is a kind of low resolution The regional model algorithm of rate.

6. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, Be characterized in that: MOS value described in step 4.2, which is segmented into, is segmented MOS value for interval with 1.

7. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, Be characterized in that: the fitting of formula described in step 4.2 refers to the form for being fitted to second order Gauss.

8. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, Be characterized in that: indices described in step 4.2 include: and variance, root mean square and determining coefficient.

9. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, Be characterized in that: the distribution situation of the delay variation in each segmentation section MOS of statistics described in step 4.3 refers to the base in step 4.2 On plinth, the distribution situation of delay variation is counted in the segmentation section MOS, mean value and standard deviation is calculated, is drawn in the form of histogram Drawing shape.

10. a kind of conversational video business QoE-QoS parameter mapping method based on statistical analysis according to claim 1, It is characterized by: the accuracy of index is suggested in verifying described in step 5.2, the specific steps are that will first calculate resulting each group of data Classify, be interval with 0.5, is classified according to MOS value；Then the suggestion index in each section MOS is replaced into original respectively The packet loss and bandwidth of beginning data, calculate MOS value；MOS value is counted in the probability in the given section MOS.