KR101462557B1 - Method of reorganizing representation in DASH - Google Patents

Abstract

A method for reorganizing representation in DASH according to the present invention includes the steps of: estimating the distribution of bandwidths of users; and reorganizing a set of the representation according to the distribution of the estimated bandwidth based on the set of the representation which is previously organized.

Description

Method of reorganizing representation in DASH [

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive streaming service technology, and more particularly, to a method of reconstructing a presentation in Dynamic Adaptive Streaming over HTTP (DASH).

Recently, the explosive increase of the network situation has made the network environment limited, and in a limited network environment, the user wants to receive continuous streaming service even if the image quality deteriorates. Therefore, if the network condition is good, high-quality streaming service is supported. However, if the network situation becomes worse, a technique of preventing the video from being broken even if image quality is low is needed.

In order to solve such a demand, the Moving Picture Experts Group (MPEG) standardizes an adaptive streaming service technology called Dynamic Adaptive Streaming over HTTP (DASH). DASH can directly predict the network status of a user from a user's device (client) using the Internet protocol (HTTP) and adaptively stream the content by requesting a proper quality content to a server in which contents are stored with various qualities Service.

In the DASH, the server prepares an image bit stream having a plurality of image quality levels, and provides the user with information on the bitstreams from the server in the HTTP format. By viewing this information, the user is provided with a media file service by requesting a URL corresponding to the media file necessary for his / her own network situation. This is a rough outline of DASH technology. One of the biggest advantages of DASH is that users can select a proper media file for themselves, so that it is possible to provide low-cost services because the resource allocation burden on the server is reduced. The user may select a lower media file to lower the quality but receive continuous video service.

In order to implement such DASH technology, a description of an XML format called MPD (Media Presentation Description) and a segment which is a data unit of the smallest unit capable of representing media data in MPD are required. Media files created with various qualities are stored in HTTP format URL addresses on the server and described in MPD. The client requests the MPD from the server, and obtains information about the corresponding file stored in the server through the MPD transmitted from the server. The user predicts the network status of the user and requests the optimum media file using the URL address using the MPD information.

Figure 1 shows the hierarchical structure of MPD in DASH. Each layer of the MPD has corresponding roles and functions. By transmitting this information to the MPD, the client can obtain information about the media file from the server. Referring to FIG. 1, an MPD is composed of one or more Periods, each Period is composed of Adaptation Sets, each Adaptation Set is composed of one or more Representations, and each Representation includes one or more segments segment. The lower level of the MPD, Period, cuts the media files in the MPD at regular intervals. The adaptation set, which is a lower level of the Period, describes the language of the content and the network bandwidth. The segment describes the quality and size of the image, and the segment includes URL information as a minimum unit for transmitting the media file. That is, MPD includes service start and end times of media files, image information, characteristics of contents, and URL information.

In the DASH system, the client (user) analyzes the MPD XML file and selects the appropriate adaptation set and presentation based on the information in the MPD. Accordingly, when the server transmits the segments of the corresponding presentation, the client stores the segment in the buffer and plays the multimedia file corresponding to the segment in the buffer. A segment is the smallest unit of data that can represent media data in a DASH system, and a multimedia file can only be played if the buffer is full of segments. That is, the multimedia data is divided into a plurality of segments. Each segment is assigned a unique URL, and has a sequence of segments, a start time, and a duration. A representation is a level that includes a segment, and each representation is classified according to the bit-rate of the image, that is, the quality of the image.

Since repetition represents the quality of the image, choosing reflination of high quality can guarantee QoS. The basic method of selecting a presentation is to allow the client to predict the bandwidth between itself and the server and select a bit-rate representation that does not exceed the expected bandwidth. In the DASH system, the number of presentations is limited due to the performance of the transcoder and the capacity of the server for real-time streaming. That is, the number of presentations that a server can provide for certain contents is determined by the performance of the server.

Previous studies in DASH have mainly focused on the user side to receive the desired bit rate representation from the server in consideration of the size of the buffer, or to make the bandwidth prediction accurate to maximize the bit rate that can be received from the server. In other words, researches to improve the QoS on the user side were mainly performed.

However, these studies were not intended to provide better quality in terms of service providers. These studies are limited to increasing user quality, since the server does not consider network changes. In existing DASH systems, service providers have been providing users with fixed video streaming quality despite changing network conditions.

However, the network conditions change depending on the actual number of users. In other words, the more users who want to receive video streaming services, the worse the network situation, and vice versa, the better the network situation. The network conditions can vary depending on, for example, a date such as a weekend or holiday, or a time of day. Therefore, in actual network systems, the bandwidth that can be requested by users varies according to the life patterns of users. For example, if users use streaming services in the early morning hours with a small number of users, the bandwidth available to each individual will be relatively large compared to the daytime hours when the number of users is large. Conversely, during the day when the number of users is high, the available bandwidth for each user will be small.

In existing DASH systems, however, the bitrates of the representations are fixed, even if the quality requested from users changes. In other words, the service provider provides the users with fixed video quality streaming despite the changes in network conditions. However, if the server considers the network situation between the server and the users, the users will be provided with better quality of service. For example, if the network situation between the server and the user is better, there will be few users who want a low bit rate representation, so DASH's representations need to support high bit rates. Conversely, if the network situation between the server and the user is bad, some of the high bitrate representations will not be used, so the DASH's representations need to support low bit rates. If the service provider provides adaptive video streaming according to the network conditions, users will be provided with enhanced QoS.

SUMMARY OF THE INVENTION Accordingly, the present invention has been made in view of the above problems, and it is an object of the present invention to provide a method of reconfiguring a presentation so that users can receive a high-quality service as a whole.

According to an aspect of the present invention, there is provided a method of reconstructing a representation in a DASH, the method comprising: (a) estimating a bandwidth distribution of users; And (b) reconstructing a set of representations according to the distribution of the estimated bandwidth, based on the set of representations already configured.

The step (b) may reconfigure a set of representations so that the bandwidth loss of the users is minimized.

The step (b) is repeatedly performed, and it may be repeated until the set of the already configured presentation and the set of the reconstructed representation are the same.

In the step (b), reconstructing the set of representations may be to determine the bitrate of each representation constituting the set of representations.

The determining of the bit rate of the k-th representation in determining the bit rate of each of the representations may include determining a bit rate of the (k-1) -th representation determined in step And the distribution of the estimated bandwidth.

The determination of the bit rate of the k-th representation may be made according to the following equation.

=0

= 0

Here, r _k , r _k-1 , and r _{k + 1} represent the bit rates of the k-th representation, the already determined k-1-th representation and the already determined k + 1 -th representation, σ denotes the mean and variance of users 'bandwidths, and p (b) denotes the probability density function of users' bandwidths.

In the step (a), the estimation of the bandwidth distribution may be performed using a Gaussian distribution defined by an average and variance of users' bandwidths.

According to an aspect of the present invention, there is provided a computer-readable recording medium storing a program for executing the above-described method of reconstructing a presentation.

According to the present invention, the presentation can be reconfigured so that users can receive high-quality services as a whole according to the network conditions.

Figure 1 shows the hierarchical structure of MPD in DASH.
2 shows an example of distribution of users and distribution of the presentation.
FIG. 3 shows an example of a reconstruction of a representation in consideration of the distribution of users.
4 shows that the distribution of users is changed due to the change of the network situation.
Figure 5 shows an example of a reconstruction of a presentation according to a change in user distribution.
FIG. 6 shows the result of user distribution analysis corresponding to FIG. 2. FIG.
FIG. 7 shows a user distribution analysis result corresponding to FIG.
8 is a view for explaining the equation (3).
9 is a diagram for explaining the equation (5).
Figure 10 shows a flow diagram of a method for reconstructing a presentation in accordance with an embodiment of the present invention.
11 is a flowchart showing the step 1020 of FIG. 10 in more detail.
FIG. 12 is a graph showing a change in the average bit rate of users as a result of simulation according to an embodiment of the present invention, by iterating reconstruction of a representation set.
13 shows a result of simulating the embodiment of the present invention with different numbers of representations.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. In the following description and the accompanying drawings, substantially the same components are denoted by the same reference numerals, and redundant description will be omitted. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

The ideal solution to ensure that all users receive high quality of service overall, depending on network conditions, is to provide the appropriate bit rate for each user on the server. However, in a real-time video streaming system, the number of representations is limited because the server's CPU capability and transcoder performance are limited. So, in reality, many users are forced to see losses. For example, assuming that the user's maximum bandwidth is 1 Mbps but the maximum bitrate of the presentation is 800 Kbps, the user will see a loss of 200 Kbps. If the server provides a 1 Mbps representation, the user will not see any loss.

In existing DASH systems, the bitrates of the representations are fixed despite the changing network conditions. However, if we reconfigure the representations according to user demand and network conditions, the average QoS of all users could be improved. For example, suppose there are five users with different bandwidths, and the server can provide two representations. Figure 2 shows the distributions of users and distributions in this case. In the example of FIG. 2, the presentation is determined regardless of the distribution of users. Since the bit rate of the presentation requested from the user is less than the bandwidth of the corresponding user, the average bit rate of all users is (128 Kbps × 4 + 320 Kbps) / 5 = 166.4 Kbps.

FIG. 3 shows a reconstruction of the presentation in consideration of the distribution of the above-mentioned users. If the presentation is reconstructed as shown in FIG. 3, the average bit rate of all users will increase to (128 Kbps + 256 Kbps 4) /5=230.4 Kbps. Thus, it can be seen that if the presentation is reconstructed according to the bandwidth distribution of the users, the average bit rate of the users can be increased.

Furthermore, the network situation may change and the distribution of the user may change, for example, as shown in FIG. In this case, the average bit rate of the users is smaller than that of FIG. 3, and the value is (128 Kbps × 2 + 256 Kbps × 3) / 5 = 204.8 Kbps. Thus, the representation needs to be reconstructed. If the representation is reconstructed as shown in Fig. 5, an average bit rate of (128 Kbps x 2 + 320 Kbps x 3) / 5 = 243.2 Kbps can be obtained.

As described above, if the presentation is reconstructed according to the bandwidth distribution of users, the average bit rate of all users can be improved.

Hereinafter, embodiments of the present invention propose a method of reconstructing a presentation to be provided by a server according to a bandwidth distribution of users. For this purpose, we first derive an optimization function for reconstruction of representation.

Many symbols and abbreviations appear in this specification, and for convenience, the symbols are defined as shown in the following table.

sign Justice N Number of users (clients) n _i Number of users who selected the i-th representation M The number of representations included in the representation set R One representation set r _i The bit rate (i = 0, 1, ..., M-1) of the i- r ^j The presentation requested by the j < th > user in R loss ^j Bandwidth Loss of jth User b ^j The bandwidth of the jth user p (b) The probability density function of bandwidth (b) CB The server's maximum channel bandwidth (Channel Bandwidth)

FIG. 6 shows the result of user distribution analysis corresponding to FIG. 2. FIG. As described above, in FIG. 2, the presentation is determined regardless of the distribution of users. FIG. 7 shows a user distribution analysis result corresponding to FIG.

Referring to FIGS. 6 and 7, it can be seen that there is a bandwidth loss for each user in both cases. However, the bandwidth loss is less in the case of Fig. 7 where the representation is reconstructed. Assuming that the bandwidth of the users is fixed, increasing the average bit rate of the users can be expressed as lowering the average bandwidth loss of the users. This can be expressed as the following equation.

In Table 1, b ^j represents the bandwidth of the j-th user, and loss ^j represents the bandwidth loss of the j-th user.

Thus, embodiments of the present invention reconfigure the presentation so that the bandwidth loss of users is minimized. The bandwidth loss of the users is determined by the bandwidth distribution and the presentation of the users. That is, the goal of re-presentation reconstruction is to minimize the bandwidth loss of each user. In order to make the optimization function for representation reconstruction a convex function, the optimization function minimizes the expectation of the square of the bandwidth loss.

로 표현될 수 있다. 즉, i 번째 레프리젠테이션을 선택한 사용자의 수 n_i를 안다면 모든 사용자들의 총 비트레이트를

로서 구할 수 있다. 결국, 사용자들의 대역폭 손실이 최소화하는 목표는 다음 수학식과 같이 표현될 수 있다. As shown in Equation (1), minimizing the expected value of the square of the bandwidth loss (hereinafter referred to as " lost squared ") means maximizing the average bit rate of users. However, there is a maximum channel bandwidth (CB) that the server can provide. If the total bitrate of the users exceeds the maximum channel bandwidth (CB) of the server, some users may not receive normal streaming services even though the presentation is reconstructed. Each user may be provided with a representation of a bit rate that is less than or equal to its maximum bandwidth so that the total bit rate of all users is

. ≪ / RTI > That is, if knowing the number n _i of users who selected the i-th representation, the total bit rate of all users

. As a result, the goal of minimizing the bandwidth loss of the users can be expressed as the following equation.

이며, 여기서

이다.

은 사용자가 수신하는 레프리젠테이션이며, 사용자는 그의 최대 대역폭을 초과하는 레프리젠테이션을 요청할 수 없는 바, 대역폭 손실은 일반적으로 레프리젠테이션에 의해 결정된다. 나아가, 대역폭 손실은 레프리젠테이션들의 구간에 따라서 정해진다. 예컨대, 레프리젠테이션 r₀와 r₁ 사이의 손실 자승의 기댓값은

이다. 따라서 손실 자승의 기댓값은 다음 수학식과 같이 표현될 수 있다.Here, it can be assumed that N >> M. That is, the number of users is much larger than the number of representations. And a probability density function of the bandwidth b of the user can be used. As mentioned in Table 1, r _i is the bit rate of i (i = 0, 1, ..., M-1) th presentation and n _i is the number of users who selected the i th presentation. The bandwidth loss at bandwidth b is

, Where

to be.

Is a representation received by the user and the user can not request a representation that exceeds its maximum bandwidth such that the bandwidth loss is generally determined by the presentation. Furthermore, the bandwidth loss is determined by the interval of the representations. For example, the expected value of the loss squared between the representations r ₀ and r ₁ is

to be. Therefore, the expected value of the loss squared can be expressed by the following equation.

Letting f (R) be the equation (3), f (R) can be expressed by the following equation. 8 is a view for explaining f (R).

In general, the number of users and the distribution of users' bandwidths or probability density functions in a server can be estimated through known network information collection techniques. Therefore, the number n _i of users who selected the i-th representation can be obtained according to the following equation. 9 is a diagram for explaining the equation (5).

In Equation (5), the integral value between r _i and r _{i + 1} is the probability value of the user distributed therebetween.

에 대응하는

은 다음 수학식과 같이 표현될 수 있다. h(R)은 라그랑주 승수법(lagrange multiplier)에서 제약 사항에 관련된다. In Equation (2)

Corresponding to

Can be expressed by the following equation. h (R) relates to constraints in the lagrange multiplier.

Therefore, the optimization function can be obtained by using the Lagrange multiplier method, and the following equation represents a formula for this.

In order to solve Equation (7), the following equation can be calculated.

To calculate Equation (8), the following calculation process can be performed.

Solving Equation (8) through the calculation process of Equation (9), the following equation can be obtained.

From the equation (6), the following equation can be obtained.

The following equations can be obtained by using Equations (7), (10) and (11).

을 만족하지 않는다면, 최적의 해는

을 만족하는 경계값(boundary) 중 하나가 된다. 즉, 모든 사용자들의 총 비트레이트는 서버의 최대 채널 대역폭(CB)을 초과하지 않도록 하는 것이 바람직하다. 이를 위한 최적의 레프리젠테이션 포인트들은

을 만족하는 것이다. If we find a representation set R that satisfies the expression (12), then R is the optimal solution. However,

, Then the optimal solution is

And a boundary value satisfying the following expression. That is, it is desirable that the total bit rate of all users does not exceed the maximum channel bandwidth (CB) of the server. The optimal presentation points for this are

.

Furthermore, one assumption will be made in the embodiment of the present invention. The probability density function of users' bandwidths is Gaussian distributed according to the central limit theorem. In an actual streaming service, there are various network access methods such as ISDN, ASDL, and wireless Internet access, and each network access method has a different bandwidth, and users within the same approach have different bandwidths. Therefore, it can be assumed that the probability density function of bandwidth of users is a Gaussian distribution. Assuming a probability density function as a Gaussian distribution, the probability density function is defined as the mean and variance of the bandwidth.

Assuming that the probability density function is a Gaussian distribution, Equation (9) can be calculated as the following equation. Hereinafter, u denotes the average of the bandwidths, and? Denotes the variance of the bandwidths.

Through the calculation process of Equation (13), the following equation can be obtained from Equation (10).

,

로 놓으면, Step 2의

는 다음과 같이 계산된다.Further, in Step 2 of Equation (13), the substitution integral can be used.

,

, Then the

Is calculated as follows.

Therefore, when Equation 14 is set to 0, the following equation is obtained.

Referring to Equation 14 or 16, when determining the bit rate r _k of the _{k- th} representation, r _k is determined by r _k-1 , r _{k + 1} and the probability density function p (b) . That is, r _k is determined between r _k-1 and r _{k + 1} . However, in the embodiment of the present invention, it is necessary to perform iteratively, not by obtaining a result of optimization by changing the representations constituting the representation set once. Because according to the changed adjacent r _k r _k-1 and _{k +} r ₁ it is because should also be changed. Thus, it is desirable that the reconstruction of the presentation is performed iteratively until the representations get a stable value.

Figure 10 shows a flow diagram of a method for reconstructing a presentation in accordance with an embodiment of the present invention.

In step 1010, the server estimates the bandwidth distribution of users (clients). As described above, the distribution of users' bandwidth or the probability density function can be estimated through known network information collection techniques. Further, assuming that the probability density function is a Gaussian distribution, the bandwidth distribution can be determined by obtaining the average and variance of the number of users and the bandwidth.

In step 1020, the server reconstructs the representation set according to the estimated bandwidth distribution in step 1010, based on the representation set already configured. As shown in the figure, step 1020 is repeatedly performed. The initialization value of the presentation set (i.e., the bit rate of each presentation) is determined, and the number of presentations constituting one presentation set is fixed I suppose. At the first execution of the iteration, it is reconstructed based on the initialization value, and at the next iteration, it is reconstructed based on the reconstructed representation set through the immediately preceding iteration. The reconstruction of the representation set determines the bitrate of the representations constituting the representation set. In the embodiment of the present invention, the bitrates of each representation are sequentially determined.

In this embodiment of the present invention, the representation is reconstructed so as to minimize the bandwidth loss determined according to the bandwidth distribution of the users and the presentation set, and the above formula 14 or 16 derived for this purpose is used.

In step 1030, the server determines whether the currently reconstructed representation set is the same as the representation set already reconstructed, i.e. reconstructed through the previous iteration. If they are the same, the representations have obtained a stable value, and thus the process ends. If not, the process returns to step 1020 and the reorganization is repeated. Depending on the embodiment, it may be terminated if the difference between the currently reconstructed representation set and the reconstructed representation set through the previous iteration is smaller than a certain level.

Furthermore, since the bandwidth distribution of users may vary according to time or situation as described above, the embodiments of the present invention may allow users to change their bandwidth distribution from time to time or periodically (e.g., at specific times of the day, The bandwidth distribution may be estimated, and the above-described processes may be performed according to the estimated distribution to change the presentation configuration.

11 is a flowchart showing the step 1020 of FIG. 10 in more detail. As described above, the number of presentations constituting the representation set is M, and thus the bit rate of each representation is represented by r _k (i = 0, 1, ..., M-1) . In an embodiment of the present invention, the lowest bit rate r _{0 of the} representation is a fixed value and the highest bit rate r _{M-1 of the} representation can not exceed the maximum channel bandwidth CB of the server .

In the first step 1021, k = 0 is set, and in step 1023, k is increased by 1. In step 1025, r _k is determined. In step 1027, steps 1023 and 1025 are repeated until k = M-1. Through this process, the bit rates of the respective representations up to r ₁ , r ₂ , ..., r _M-1 are determined.

In this embodiment, the determination of the bit rate r _k of the _{k-th representation} is made according to Equation (14) to 0 or Equation (16). Therefore, as described above, r _k is determined by r _k-1 , r _{k + 1} and the probability density function p (b). Here, r _k-1 , r _{k + 1} are the determined bit rates of the already determined k-1 th presentation and the already determined k + 1 th presentation. More specifically, r _{k + 1} is the bit rate determined through the previous iteration of step 1020, and r _k-1 is the bit rate determined through the current iteration of step 1020.

In the first step 1025, a second presentation r ₁ is determined based on the first presentation r ₀ and the third presentation r ₂ . In the last step 1025, the last representation r _M-1 is determined based on r _M-2 and r _{M. In} the embodiment of the present invention, the value of r _M for determination of r _M-1 is set as the maximum channel The bandwidth CB can be set. However, if the maximum value of the bit rate is to be a value smaller than the maximum channel bandwidth (CB) of the server, the value of r _M may be set as the value.

FIG. 12 is a graph showing a change in the average bit rate of users as a result of simulation according to an embodiment of the present invention, by iterating reconstruction of a representation set. Referring to FIG. 12, it can be seen that the bit rate has entered the stabilization phase by repeating about 6-7 times. At this time, the bit rate is about 940Kbps, and it can be confirmed that the improvement is over 100Kbps in comparison with the case of using the set of initially given presentation.

Furthermore, in general, as the number of presentations increases, the average bit rate of the users can be improved. As the number of presentations increases, the present invention exhibits more excellent performance. 13 shows a result of simulating the embodiment of the present invention with different numbers of representations. Referring to FIG. 13, the bit rate increases sharply while the number of representations increases from 5 to 15, and it is seen that the bit rate increases more slowly when the number of representations increases from about 25 to 15. When the embodiment of the present invention is applied by changing the number of representations, the optimal number of presentations can be determined in consideration of the performance of the server based on the bit rate desired by the user.

The above-described embodiments of the present invention can be embodied in a general-purpose digital computer that can be embodied as a program that can be executed by a computer and operates the program using a computer-readable recording medium. The computer-readable recording medium includes a storage medium such as a magnetic storage medium (e.g., ROM, floppy disk, hard disk, etc.), optical reading medium (e.g., CD ROM,

The present invention has been described with reference to the preferred embodiments. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is defined by the appended claims rather than by the foregoing description, and all differences within the scope of equivalents thereof should be construed as being included in the present invention.

Claims (8)

A method of reconstructing a representation in a DASH,
(a) estimating a distribution of users bandwidth; And
(b) reconstructing a set of representations according to the distribution of the estimated bandwidths, based on a set of representations already configured, wherein the bits of each representation constituting the set of representations of the reconstructed representations Determining a rate using a bit rate of a neighboring sequence of presentations already determined in the already set of presentations and reconstructing the set of presentations. Reconstruction method. The method according to claim 1,
The step (b)
Wherein the set of representations is reconstructed such that a bandwidth loss of users determined according to the estimated bandwidth distribution and a set of representations is minimized. The method according to claim 1,
Wherein the step (b) is repeatedly performed until the set of the already configured presentation is the same as the set of the reconstructed representation. delete The method according to claim 1,
The determining of the bit rate of the k-th representation in determining the bit rate of each of the representations may include determining a bit rate of the (k-1) -th representation determined in step And a distribution of the estimated bandwidth. 6. The method of claim 5,
Wherein the determination of the bit rate of the k-th representation is performed according to the following equation.

= 0
Here, r _k , r _k-1 , and r _{k + 1} represent the bit rates of the k-th representation, the already determined k-1-th representation and the already determined k + 1 -th representation, σ denotes the mean and variance of users 'bandwidths, and p (b) denotes the probability density function of users' bandwidths. The method according to claim 1,
Wherein the estimation of the bandwidth distribution in step (a) is performed by a Gaussian distribution defined by an average and variance of bandwidths of users. A computer-readable recording medium having recorded thereon a program for executing the reproduction reconstruction method according to any one of claims 1 to 3 and 5 to 7.

Images