Video playback device, video playback method, and program

Abstract

A video playback device comprising a processor and a memory, the memory including instructions for causing the processor to execute the steps of: predicting future throughput using past throughput values; setting a search space for finding a quality adaptation schedule; extracting a plurality of paths in the search space; estimating a quality of experience value for each path among the extracted plurality of paths based on the future throughput; identifying the path corresponding to the highest quality of experience value among the estimated quality of experience values ​​as the quality adaptation schedule; determining a quality value of a chunk to be requested in the next future unit time interval of the current unit time interval based on the quality adaptation schedule; requesting a chunk of the determined quality value; and receiving a chunk of the determined quality value.

Description

Technical Field

[0001] The present invention relates to a video playback device, a video playback method, and a program.

Background Art

[0002] It is difficult to stream video to a large number of viewers. Each user accesses the Internet with different capabilities that involve different throughputs, which also vary as a function of time. Therefore, in order to stream video to all users, it is necessary to use different video representations (videos encoded with various resolutions, bitrates, and frame rate values) to make the video bitrate requirements adaptable to the user's available throughput. Such a method of adapting quality to the available throughput is called adaptive bitrate video streaming.

[0003] In the process of adaptive bitrate (ABR) streaming, different video representations with different throughput requirements are presented to the video playback device. The entire video is divided into a plurality of segments called chunks, and it is possible to adjust the quality of the video in units of chunks. The video playback device performs a process of providing the highest perceived quality of experience (QoE) while avoiding buffer depletion that causes frequent quality changes and video pauses by selecting an appropriate sequence of chunks to maximize the encoding quality. To achieve this goal, it is necessary to consider the following two main factors: First, it is necessary to define appropriate rules for the chunk request determination mechanism. Second, since chunk selection may depend on throughput estimation, an accurate throughput prediction is required for appropriate chunk selection.

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

Non-Patent Document 4

Non-Patent Document 5

Non-Patent Document 6

Summary of the Invention

Problems to be Solved by the Invention

[0005] Regarding the selection rules of chunk packets, many studies have been conducted. However, in these studies, the time scale considered in chunk selection is limited. In fact, these studies focus on selecting the next chunk in relation to the problem of "which chunk should be requested next" that the video playback device has.

[0006] However, from such a short-term perspective, it may not be possible to predict a decrease in the efficiency of network access, and only a response to the observed event of efficiency decrease can be obtained. As a result, the resulting QoE may not be optimal.

[0007] As an example, consider the case where a user watches a video in a vehicle. At a certain time, the user can access a high-quality network, enabling the user to watch a high-quality video. However, when the vehicle enters a tunnel, the available throughput drops significantly. In this situation, a conventional video playback device with a "short-term perspective" can do nothing but significantly reduce the video quality to avoid stopping video provision when the device measures the drop in throughput, and no other options are provided.

[0008] There is a need for a video playback technology that can improve the perceived quality of the user in a difficult network environment.

Means for Solving the Problem

[0009] According to one aspect of the present invention, there is provided a video playback device including a transmission unit, a reception unit, a processor, and a memory. The memory causes the processor to perform steps of predicting future throughput during a future first time interval including m future unit time intervals using past throughput values; setting a search space defined by k quality levels and a second time interval including the current unit time interval, the m future unit time intervals, and n past unit time intervals for finding a quality adaptation schedule; extracting, from the search space, a plurality of paths, each path of the plurality of paths being obtained by selecting any one of the k quality levels for each unit time interval included in the second time interval; estimating a perceived quality value for each of the extracted plurality of paths based on the future throughput; specifying, as the quality adaptation schedule, a path corresponding to the highest perceived quality value among the estimated perceived quality values; determining a quality value of a chunk to be requested for a future unit time interval next to the current unit time interval based on the quality adaptation schedule; requesting, via the transmission unit, a chunk having the determined quality value; and receiving, via the reception unit, the chunk having the determined quality value.

Effect of the Invention

[0010] According to an embodiment, a video playback technology is provided that enables improving the perceived quality of a user in a difficult network environment.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Embodiments for Carrying Out the Invention

[0012] Hereinafter, the outlines of different algorithms will be introduced. These can be classified into two main categories: the rule-based ABR model and the learning-based ABR model. In the first category, the decision-making process for chunk selection is based on handcrafted rules, while in the second category, reinforcement learning is used to determine the chunks to be requested.

[0013] There are two main types of models in the rule-based ABR model: a model that uses only buffer filling rate information to make chunk requests and an algorithm that uses throughput prediction. Regarding the method based only on the buffer, in Non-Patent Document 1, a method of requesting chunks depending on a linear function of the buffer filling rate has been proposed. When the buffer is full, higher-quality chunks are requested. On the other hand, when the filling rate of the buffer is low, low-quality video is requested to avoid video stoppage. In such a method, a relaxation factor for the case where only a high-quality level is requested when the buffer filling rate reaches the lowest level is considered, thereby avoiding buffer depletion caused by requesting chunks of too high quality.

[0014] Since there is a non - linear relationship between the bitrate and quality of a video, using a linear relationship between the buffer filling rate and the selected chunks may not give an optimal solution. Therefore, the BOLA algorithm has been proposed. In this algorithm, chunks are selected based on a threshold of the buffer filling rate. When the buffer filling rate reaches a certain filling rate level, a predetermined quality level is selected. The process of determining the threshold is performed for each video content using information on the video bitrates and corresponding qualities of different video representations, which are video bitrates and corresponding qualities that allow the video to be utilized. In previously introduced methods, only the buffer filling rate is considered to select chunks. However, methods that also consider throughput information are being investigated. The idea behind these methods is to obtain an estimated value of future throughput using knowledge about the chunks downloaded in the past, and to select chunks based on the prediction of the ability of the video playback device to download chunks within time, taking into account the available network conditions.

[0015] The simplest method is an algorithm based only on rate. In this method, chunks are selected such that the bitrate required for the selected chunks is the bitrate just below the estimated throughput. However, in past research, it has been shown that users prefer a more stable quality than a fluctuating quality. So, one of the limitations of such a method is that it may cause frequent quality changes and potentially reduce the QoE. Therefore, a model predictive control (MPC) approach has been proposed, taking into account the effects of quality fluctuations and buffer information (to consider video pauses) (Non - Patent Document 2).

[0016] The general idea of such a type of ABR control mechanism is to optimize the QoE evaluation function given by the linear addition of three factors related to QoE: the quality of the requested chunks, the smoothness of the quality change between consecutive chunks, and the video stop and initial load delay period. Then, the future throughput is estimated, and chunks are selected to maximize the quality and minimize the quality change and video stop.

[0017] One of the challenges of such a method is the ability to predict future throughput values. To address this problem, different models derived from the underlying MPC algorithm have been proposed. In Non-Patent Document 2, throughput prediction is performed using the harmonic mean. Next, in Non-Patent Document 2, the authors provide an alternative approach called RobustMPC, which can reduce the overestimation of throughput by weighting the past throughput prediction error with respect to the predicted throughput value.

[0018] In Non-Patent Document 3, an algorithm called Fugu has been proposed, and a deep neural network (DNN) is used to predict the transmission time. This algorithm uses features such as the past chunk size, past transmission time, Transmission Control Protocol (TCP) statistics (congestion window, packets without acknowledgment, round-trip time (RTT), TCP estimated throughput), etc. to estimate the time required to download chunks of different qualities. Based on these estimated values, it is possible to adapt the MPC rule weights related to quality, quality variation, and events related to video stop for the determination of the chunks to be selected to provide the best experience for the user.

[0019] BayesMPC proposes to model throughput using a learned Bayes neural network. According to this approach, the distribution of throughput values is predicted, enabling a probabilistic view of the chunk selection algorithm while using the underlying MPC rules. Finally, the Hidden Markov Model (HMM) is also proposed as a way to perform throughput prediction.

[0020] Instead of handcrafted rule-based ABR models, previous studies have considered using reinforcement learning for the chunk selection mechanism. Pensive (Non-Patent Document 4) is a guideline for this type of model. The general idea of this algorithm is that rules such as those defined for MPC can be considered arbitrary and can be subject to human error when designing those rules.

[0021] For example, the weighting between requiring high-quality chunks and aiming to maintain a certain quality is left as a free parameter, and it may not be an easy task to set this value. Furthermore, the characteristics of video stalls are considered only from the perspective of the duration of video stalls linearly combined with characteristics related to encoding quality, and there is room for discussion about this as well.

[0022] Therefore, instead of defining handcrafted rules for the chunk selection process, it is proposed to use a data-driven approach to learn which rules should be used to determine the chunks to be requested.

[0023] Next, Pensive (Non-Patent Document 4) uses information about buffer filling speed, data on throughput experienced in the past, and the sizes of chunks requested in the past to directly predict the chunks to be selected. However, it is worth mentioning that the training was based on the loss using the objective function of MPC. Although Pensive (Non-Patent Document 4) does not rely on throughput prediction, the authors proposed an extension and a new model that utilizes throughput prediction information to improve the performance of the algorithm.

[0024] Next, further extensions of these types of models are being considered by using adversarial generative networks (GANs). An example of this type of model is the Tiyuntsong model. The idea of this type of model is to use the learned loss instead of relying on a handcrafted loss function that was previously based on the objective function of MPC.

[0025] In the following, embodiments of the present invention will be described with reference to the drawings.

[0026] The adaptation of the video bitrate and quality to the available throughput is processed by the adaptive bitrate control of the video playback device.

[0027] The ABR control mechanism depends on selecting chunks according to handcrafted rules or learning-based rules. Here, a chunk may be a part of a video encoded at a predetermined bitrate.

[0028] Many studies have been conducted on the selection rules of chunk packets. However, in these studies, the time scales considered in chunk selection are limited. In fact, these studies focus on selecting the next chunk in relation to the problem of "which chunk should be requested next" that the video playback device has.

[0029] However, from such a short-term perspective, it is impossible to predict the efficiency degradation of network access, and only the response to the observed efficiency degradation event can be obtained. As a result, the resulting QoE may not be optimal.

[0030] As an example, consider the case where a user watches a video in a vehicle. At a certain time, the user can perform high-quality network access, and thus the user can watch high-quality video. However, when the vehicle enters a tunnel, the available throughput drops significantly.

[0031] In this situation, a video playback device with a "short-term perspective" can only reduce the video quality significantly to avoid stopping video provision when the device measures the drop in throughput, and no other options are provided.

[0032] Executing such a short-term chunk selection mechanism will narrow down the options for other quality adaptation under difficult network conditions, which is problematic. To solve the above problems, it is proposed to perform chunk selection based on long-term prediction of throughput.

[0033] For example, considering the above example of the vehicle entering the tunnel, the gist of the proposed method is that if the video playback device can predict a significant drop in throughput in advance, it can find solutions other than significantly reducing the video quality.

[0034] For example, before the throughput drops, the video playback device can use the available high throughput to capture the video into the buffer at a faster speed.

[0035] This will ultimately be achieved at the cost of slightly reducing the quality. However, as a result, while the vehicle is in the tunnel with limited network access, the user can watch high-quality video.

[0036] By executing chunk selection based on long-term throughput prediction, higher quality and more stable quality are provided to the user, and the perceived quality of the user in a difficult network environment is improved.

[0037] The proposed algorithm is executed in the device and aims to define the procedure for long-term chunk request scheduling.

[0038] Similar to the ABR control method, the proposed algorithm takes as input a Media Presentation Description (MPD) that gives information about multiple quality levels available for the video.

[0039] Figure 1 shows the key steps of the proposed method. The key steps are as follows.

[0040] 1) Estimate the temporal variation of throughput for a time interval of up to T seconds (the value of T will be explained below).

[0041] 2) Find the optimal chunk scheduling for a time interval of T seconds (Figure 2).

[0042] 3) Request the first chunk of the found optimal path (optimal scheduling).

[0043] Thereafter, the algorithm returns to step 1 and estimates again the temporal variation of throughput for a period of up to T seconds.

[0044] The algorithm loops through this process until it reaches the end of video playback or the user finishes watching the video.

[0045] FIG. 3 is a diagram showing a hardware configuration example of the video playback device 10 according to an embodiment of the present invention. The video playback device 10 in FIG. 3 includes a drive device 100, an auxiliary storage device 102, a memory device 103, a central processing unit (CPU) 104, and an interface device 105, etc., which are mutually connected by a bus B.

[0046] The program for realizing the processing in the video playback device 10 is provided by a recording medium 101 such as a CD-ROM. When the recording medium 101 storing the program is set in the drive device 100, the program is installed from the recording medium 101 to the auxiliary storage device 102 via the drive device 100. However, the installation of the program does not necessarily have to be performed from the recording medium 101, and it may be downloaded from another computer via a network. The auxiliary storage device 102 stores the installed program and also stores necessary files, data, etc.

[0047] When an activation instruction for the program is given, the memory device 103 reads out and stores the program from the auxiliary storage device 102. The CPU 104 executes the functions related to the video playback device 10 according to the program stored in the memory device 103. The interface device 105 is used as an interface for connecting to a network.

[0048] FIG. 4 is a diagram showing a functional configuration example of the video playback device 10 according to an embodiment of the present invention. In FIG. 4, the video playback device 10 includes a throughput prediction unit 11, a bit rate control unit 12, a communication unit 13, and a buffer 14. These units are realized by the processing executed by the CPU 104 for one or more programs installed in the video playback device 10. That is, these units are realized by the cooperation of the hardware resources of the video playback device 10 and the programs (software) installed in the video playback device 10. Note that the communication unit 13 may include a transmission device and a reception device.

[0049] Hereinafter, an example of the processing procedure executed by the video playback device 10 will be described. FIG. 5 is a flowchart for explaining an example of the processing procedure executed by the video playback device 10.

[0050] Before starting the optimization procedure, it is necessary to determine the value of T. T is a parameter determined by the user. The larger the value of T, the higher the performance. However, if the value of T is increased, the calculation cost increases. Therefore, it is necessary to resolve the trade-off between the calculation cost and the quality level. In the simulation, when the value of T is set to 20 seconds, a significant improvement in quality has been found, but other values can also be considered.

[0051] In step S101 of FIG. 5, the throughput prediction unit 11 predicts the future throughput for the next T seconds. Predicting throughput is a difficult task and has received much attention.

[0052] It is possible to predict the future throughput by applying any of different techniques such as the arithmetic mean, harmonic mean, deep neural network, or hidden Markov model.

[0053] However, in these methods, only the immediate throughput is predicted. Therefore, the throughput prediction unit 11 uses time series prediction techniques to predict the throughput for a longer time frame.

[0054] As prediction methods used by the throughput prediction unit 11, different methods can be considered, such as the use of autoregressive models such as autoregressive integrated moving average (ARIMA) or advanced learning-based techniques.

[0055] Hereinafter, as a prediction method used by the throughput prediction unit 11, the "transformer-based deep neural network throughput prediction algorithm" (Non-Patent Document 5) is proposed, but alternative models can also be considered.

[0056] When using this method, the throughput prediction unit 11 predicts the future throughput for the next T seconds using past throughput values.

[0057] In step S102 of FIG. 5, the bitrate control unit 12 sets a search space for finding an optimal quality adaptation schedule at time intervals of the next T seconds.

[0058] FIG. 2 is a diagram showing an example of a search space for finding an optimal quality adaptation schedule. In the example of FIG. 2, the horizontal axis of the search space is the time axis, and the vertical axis represents the video quality level.

[0059] In the example of FIG. 2, in addition to the current time t and t+1, t+2,..., t+4 corresponding to the time interval of the next T seconds, the past times t-2 and t-1 are included in the search space. Also, as video quality levels, six different quality levels from 144p to 1080p maximum are available. That is, as shown in the example of FIG. 2, the bitrate control unit 12 sets the search space by setting different quality levels (from 144p to 1080p maximum in the example of FIG. 2) as the vertical axis for each unit time (T / 4 in the example of FIG. 2). Note that in the example of FIG. 2, the next T seconds are divided into four unit times, but the embodiments are not limited to this example. The next T seconds may be divided into more than four unit times, or the next T seconds may be divided into less than four unit times. Also, the quality levels are not limited to six levels, and may be more than six levels or less than six levels.

[0060] Note that in the example of FIG. 2, the line segment from t-2 to t indicates the previous chunk selected by the bitrate control unit 12. At time t, the bitrate control unit 12 needs to select the quality level of the chunk for the next unit time t+1. For this purpose, in step S103 of FIG. 5, the bitrate control unit 12 selects different paths in the search space corresponding to different quality adaptation schedules in the next T seconds as the evaluation targets of QoE.

[0061] In the example of FIG. 2, three paths are shown that are expected to provide QoE1, QoE2, and QoE3 as different paths in the search space.

[0062] In step S104 of FIG. 5, the bitrate control unit 12 uses the future throughput predicted in step S101 and the QoE model to evaluate the QoE for a specific path. As the QoE model, it is possible to use the model described in Non-Patent Document 6. However, it is also possible to use other QoE models described in ITU-T Rec. P.1203 or ITU-T Rec. P.1204.

[0063] These standardized models can be utilized in different modes, using different types of information ranging from low-complexity models that use only parameters such as bitrate, frame rate, and resolution to high-complexity models that use even bitstream or pixel information.

[0064] The bitrate control unit 12 predicts the QoE of a specific path based on the throughput in the past P seconds and the subsequent T seconds.

[0065] By considering the chunks downloaded in the past, it is possible to address the temporal smoothness of the quality.

[0066] In the example shown in FIG. 2, the time interval of one chunk is set to 1 second, P is set to 2 seconds, and T is set to 4 seconds.

[0067] Regarding the request for future chunks, although three different paths are shown in FIG. 2, in order to be able to find the path that gives the highest QoE, an evaluation regarding QoE should be performed considering the case where all qualities among the available multiple quality levels are applied.

[0068] The process of calculating the QoE of a specific path is performed as follows: Each chunk in the path corresponds to different quality levels, characterized by resolution r, video bitrate b v , frame rate f, and audio bitrate b a .

[0069] By using the characteristics of each chunk in a specific path, the value per second of each parameter among these multiple quality-related parameters can be obtained.

[0070] Next, based on the QoE model of Non-Patent Document 6, the video quality per second is evaluated using Equation 1, the audio quality per second is evaluated using Equation 4, and the video-audio quality per second is evaluated using Equation 5.

[0071]

Equation

[0072]

Equation

[0073]

Equation

[0074] Note that in Equation 1, X and Y are defined by Equation 2 and Equation 3.

[0075]

Equation

[0076]

Equation

[0077] The time aggregation of the score related to the encoding quality per second into the score per path is given by Equations 6 to 9.

[0078]

Number

[0079]

Number

[0080]

Number

[0081]

Number

[0082] Finally, Equation 10 is used to take into account the impact of the stop of video provision.

[0083]

Number

[0084] In Equation 10, n b is the number of stops of video provision that occurred on the path to be considered, t b is the total time interval of the stop of video provision, and a b is the average of the time intervals between the events of the stop of video provision.

[0085] In these equations, v 1 -v 7 , a 1 -a 3 , m 1 -m 4 , t 1 -t 5 and s 1 -s 3 are model parameters obtained by using regression with a subjective experimental database.

[0086] If L is the number of quality levels and N is the number of chunks to be downloaded in the next T seconds, then L N It is necessary to compare the QoE of different paths of the book.

[0087] By applying restrictions to the evaluation of quality adaptation considering the computational cost, it is possible to reduce the number of paths to be evaluated.

[0088] As an optimization candidate, only an increase or decrease in quality by only the A level may be allowed.

[0089] For example, in the example of FIG. 2, A = 1, and at time t, the quality level of the previously requested chunk is 360p.

[0090] At time t, it is considered possible to request only quality levels 480p, 360p, and 240p.

[0091] At time t + 2, depending on the selection made at t + 1, the quality value can range from 720p to 144p.

[0092] By performing such optimization, it leads to smooth quality adaptation that results in a higher QoE, L N While obtaining the effect of forcing smooth quality adaptation, which would be selected even if a full search for all paths of the book is performed, the search space can be reduced to (2 × A + 1) N paths of the book.

[0093] In step S105 of FIG. 5, the bitrate control unit 12 determines L among the N paths of the book and selects the path that is expected to provide the highest QoE. As described above, when only an increase or decrease in quality by only the A level is allowed as an optimization candidate, the bitrate control unit 12 selects (2 × A + 1) NSelect the path that is expected to provide the highest QoE among the book's paths.

[0094] In the case of starting a video, which is a special case, since past throughput data cannot be obtained, it may not be possible to predict the future throughput for the next T seconds. In this case, different alternatives can be considered.

[0095] 1) The bitrate control unit 12 may choose to start at the lowest quality level. In this case, a fast initial load is possible.

[0096] 2) The bitrate control unit 12 may allow the user to select the initial quality value.

[0097] 3) The bitrate control unit 12 may estimate the future throughput based on the previously viewed videos by the user.

[0098] For example, when the user finishes viewing a video and switches to a new video, information about the available throughput measured in the previous video can be used to predict the throughput of the next video, thereby enabling the video to be started at an appropriate quality level.

[0099] Similarly, by using information such as the throughput measured at the stage of loading the web interface as a clue, it is also possible to predict the user's available throughput and select an appropriate initial quality level.

[0100] In step S106 of FIG. 5, the bitrate control unit 12 makes a chunk request via the communication unit 13.

[0101] In step S105, after selecting the path that is expected to provide the highest QoE, in step S106, the bitrate control unit 12 requests, via the communication unit 13, a chunk of the quality level corresponding to the immediately following chunk included in the selected path.

[0102] For example, in FIG. 2, assume that the path that brings the QoE value QoE 2 provides the highest QoE to the user.

[0103] In this case, the bitrate control unit 12 requests, via the communication unit 13, a chunk corresponding to the quality level 360p.

[0104] Thereafter, the communication unit 13 acquires, via a network such as the Internet, the chunk of the quality level requested in step S106 and stores it in the buffer 14. As a result, the video playback device 10 can provide the user with a video of higher quality and more stable quality.

[0105] In step S107 of FIG. 5, the bitrate control unit 12 determines whether the end of video playback has been reached or whether the user has ended viewing the video.

[0106] In step S107, if it is determined that the end of video playback has been reached or if it is determined that the user has ended viewing the video (S107: YES), the bitrate control unit 12 ends the processing procedure.

[0107] In step S107, if it is determined that the end of video playback has not been reached and it is determined that the user has not ended viewing the video (S107: NO), the processing procedure returns to S101, and the throughput prediction unit 11 uses the throughput measured from chunk t-1 to t+1 to predict the throughput from t+2 to t+5.

[0108] In steps S102 to S105, the bitrate control unit 12 recalculates the optimal path between t+2 and t+5 based on the estimated throughput and selects the chunk to be downloaded at t+2.

[0109] This procedure is repeated in a loop until the user stops viewing or the video ends.

[0110] The above embodiments show how to use the prediction of long-term throughput by time series analysis in an adaptive bitrate control method that performs long-term chunk request scheduling. According to this method, it is possible to evolve a conventional narrow-view video playback device into a video playback device that can predict throughput fluctuations and provide higher quality to users.

[0111] Also, the above embodiments show that it is possible to use the latest perceived quality model to evaluate the quality of a specific chunk selection path. By using the knowledge of the QoE model that defines an appropriate quality adaptation procedure, the perceived quality of the user is improved by executing the procedure.

[0112] Also, the above embodiments show that in order to reduce the computational complexity, the quality adaptation is limited to an increase or decrease that does not exceed a user-defined threshold. Thereby, while significantly reducing the computational cost, it is possible to enforce the temporal smoothness of the quality.

[0113] The above embodiments describe a method of using long-term throughput prediction to determine the path of the chunk request schedule in order to improve the overall perceived experience when a user views a video in a network where the throughput varies. Different from the conventional method of determining which chunk should be requested for each chunk, the proposed method estimates the future throughput in the long term and estimates which chunk scheduling path gives the highest perceived experience to the user. Considering the new measurement of throughput, the specified path is updated periodically, enabling continuous update of the chunk request scheduling path.

[0114] Although the embodiments of the present invention have been described in detail above, the present invention is not limited to such specific embodiments, and various modifications and changes are possible within the scope of the gist of the present invention described in the claims.

Claims

1. A video playback device, comprising: a transmission unit; a reception unit; a processor; and a memory, wherein the memory causes the processor to: predict future throughput during a future first time interval including m future unit time intervals using past throughput values; set a search space defined by k quality levels and a second time interval including a current unit time interval, the m future unit time intervals, and n past unit time intervals, for finding a quality adaptation schedule; extract, from the search space, a plurality of paths, each path in the plurality of paths being obtained by selecting any one of the k quality levels for each unit time interval included in the second time interval; for each path among the plurality of extracted paths, estimate a perceived quality value based on the future throughput; specify, as the quality adaptation schedule, the path corresponding to the highest perceived quality value among the estimated perceived quality values; determine, based on the quality adaptation schedule, a quality value of a chunk required for the next future unit time interval of the current unit time interval; request, via the transmission unit, a chunk having the determined quality value; and receive, via the reception unit, the chunk having the determined quality value, and includes instructions for causing execution of the above steps, a video playback device.

2. The video playback device according to claim 1. The step of extracting allows only differences below level A as the difference in quality levels in two adjacent unit time intervals, so that in the future m unit time intervals, (2 × A + 1) ｍ extracts the path of the book

3. The step of predicting the future throughput uses the throughput measured in the n past unit time intervals to predict the throughput of the m future unit time intervals. The video playback device according to claim 1.

4. In the step of predicting the future throughput, when the video starts, the throughput of another video previously viewed or a preset low throughput is applied as the future throughput. The video playback device according to claim 1.

5. The step of estimating the perceived quality value evaluates the perceived quality value using a predetermined perceived quality model. The video playback device according to claim 1.

6. A video playback method executed by a computer, comprising: ​ Predicting future throughput during a future first time interval including m future unit time intervals using past throughput values; Setting a search space defined by k quality levels and a second time interval including the current unit time interval, the m future unit time intervals, and n past unit time intervals, the search space for finding a quality adaptation schedule; Extracting, in the search space, a plurality of paths, each path of the plurality of paths being obtained by selecting any one of the k quality levels for each unit time interval included in the second time interval; Estimating a perceived quality value for each path of the plurality of extracted paths based on the future throughput; Identifying, as the quality adaptation schedule, the path corresponding to the highest perceived quality value among the estimated perceived quality values; Determining, based on the quality adaptation schedule, the quality value of a chunk required for the next future unit time interval of the current unit time interval; Requesting the determined chunk of quality values; and Receiving the determined chunk of quality values, A video playback method comprising the steps of: Claim 7 A non-transitory computer-readable storage medium storing a video playback program, wherein when the video playback program is executed by a computer, the computer Predicting future throughput during a future first time interval including m future unit time intervals using past throughput values; Setting a search space defined by k quality levels and a second time interval including the current unit time interval, the m future unit time intervals, and n past unit time intervals, the search space for finding a quality adaptation schedule; Extracting, in the search space, a plurality of paths, each path of the plurality of paths being obtained by selecting any one of the k quality levels for each unit time interval included in the second time interval; Estimating a perceived quality value for each path of the plurality of extracted paths based on the future throughput; Specifying, as a schedule for quality adaptation, a path corresponding to the highest perceived quality value among the estimated perceived quality values; Determining a quality value of a chunk required for the next future unit time interval of the current unit time interval based on the schedule for quality adaptation; Requesting a chunk of the determined quality value; and Receiving a chunk of the determined quality value, A non-transitory computer-readable storage medium that executes the above steps.

Images