WO2023024316A1 - 一种基于网络健康指数的视频质量评估方法与系统 - Google Patents

一种基于网络健康指数的视频质量评估方法与系统 Download PDF

Info

Publication number
WO2023024316A1
WO2023024316A1 PCT/CN2021/135674 CN2021135674W WO2023024316A1 WO 2023024316 A1 WO2023024316 A1 WO 2023024316A1 CN 2021135674 W CN2021135674 W CN 2021135674W WO 2023024316 A1 WO2023024316 A1 WO 2023024316A1
Authority
WO
WIPO (PCT)
Prior art keywords
health index
video
network health
network
freeze
Prior art date
Application number
PCT/CN2021/135674
Other languages
English (en)
French (fr)
Inventor
丁华
奚溪
魏峥
施唯佳
张铮凯
Original Assignee
天翼数字生活科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 天翼数字生活科技有限公司 filed Critical 天翼数字生活科技有限公司
Publication of WO2023024316A1 publication Critical patent/WO2023024316A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N17/00—Diagnosis, testing or measuring for television systems or their details
    • H04N17/004—Diagnosis, testing or measuring for television systems or their details for digital television systems
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N17/00—Diagnosis, testing or measuring for television systems or their details
    • H04N17/02—Diagnosis, testing or measuring for television systems or their details for colour television signals

Definitions

  • the invention relates to the Internet field, in particular to a video quality evaluation method and system based on a network health index for real-time streams such as IPTV.
  • fluency is the most important item in the video quality experience index.
  • the evaluation of video quality mainly includes subjective quality evaluation and objective quality evaluation, and can also be combined subjectively and objectively.
  • Subjective quality assessment usually needs to be carried out under restricted environmental conditions. After watching in groups, people will score subjectively according to grades. It is difficult to implement directly using the experience results of human eyes, and it is not convenient to evaluate a large number of videos, and individual scoring is due to subjective There is a large deviation due to factors.
  • the subjective-objective combination method combines the objective quality assessment with the individual average score, and the fitting effect is not good, and there is also a large deviation due to the subjective factors of the individual score.
  • objective quality assessment can be further divided into three categories: full reference, partial reference, and no reference:
  • Partial reference the method is to compare part of the original video and apply calculations, which is easy to implement and can reduce bandwidth overhead compared to full reference, but the evaluation value is not necessarily consistent with the human eye experience, and there is a certain deviation;
  • the method includes analyzing video frame information, analyzing network transmission data, and analyzing chip freeze data.
  • the degree of consistency between the evaluation result and the subjective experience of the human eye is the most important indicator. Based on this, it can be seen from the above-mentioned quality assessment methods that only the full-reference type and the no-reference type for analyzing stuck data can achieve the same effect as the human eye experience, but the full-reference type is used because the original video is not easy to obtain.
  • the scope is limited. Therefore, only the latter, that is, based on the frame information provided by the chip, can analyze the frame information provided by the chip, and the objective quality assessment without reference type can better evaluate the video quality.
  • set-top box A has chip frame frame information.
  • Statistical module based on the frame information provided by the chip, analyzes the freeze data to better realize the evaluation of video quality.
  • the set-top boxes in the northern provinces have no hardware chip frame freeze information statistics module due to player and hardware problems, or there is no freeze data in the hardware chip, or the player software has freeze data although the hardware chip has freeze data
  • the stall data is not transmitted to the probe, so the above-mentioned objective quality assessment without reference based on the frame information provided by the chip to analyze the stall data cannot be realized.
  • those set-top boxes in the entire network that do not have a statistical module for chip frame jam information will make it difficult to conduct network-wide statistical video quality evaluation.
  • the quality assessment method solves the problems of network-wide quality monitoring and unified evaluation standards for the northern and southern provinces.
  • the present invention proposes a network health index algorithm embedded in the set-top box probe, obtains the MDI (Media Delivery Index) value of the media transmission coefficient, and optimizes and corrects the algorithm coefficient by combining the network index and the stall curve fitting, so as to better Simulate and analyze the evaluation method of stuck data, and calculate and evaluate the network health index.
  • the technical scheme of the present invention is less difficult to implement and has high accuracy. There is no need for the entire network set-top box to adapt to the complex chip frame information stall information library, so that the video quality rate can be correctly evaluated in the scene where the player cannot provide chip information.
  • the video quality assessment system of the present invention includes: a video platform; a set-top box with a chip frame freeze information statistics module; and a set-top box without a chip frame freeze information statistics module but implanted with a video quality assessment algorithm module based on the network health index h_mos set top box.
  • the h_mos-based video quality assessment algorithm module is obtained by fitting the video quality assessment based on the MDI calculation h_mos with the video quality assessment based on the frame freeze information analysis freeze data provided by the chip.
  • the video quality evaluation method of the present invention comprises the following steps: in the test set-top box having the algorithm module based on h_mos and the evaluation module based on the chip stuck at the same time being located in the laboratory, the video quality evaluation based on the MDI calculation h_mos and the video quality evaluation based on the chip provided Frame freeze information analysis and video quality assessment of freeze data for fitting, to obtain an ideal algorithm module for h_mos-based evaluation of the evaluation effect of simulated analysis freeze data; implant the ideal algorithm module into a set-top box that lacks the ability to analyze freeze data ;
  • the set-top box reports h_mos to the video platform once at a fixed time interval, and if the reported h_mos is lower than the critical threshold of h_mos, a bad quality block is recorded, and the set-top box with a cumulative number of bad quality blocks higher than a certain percentage throughout the day is listed as a bad quality user ;Based on the number of users with poor quality and the total number of users who have played videos in the whole day, the scale of
  • the curve of the relationship between the h_mos data value and the network damage and packet loss control is fitted to be similar to the curve between the data value of the number of freeze times and the network damage and packet loss control; case as the h_mos critical threshold.
  • FIG. 1 are the overall module schematic diagram of existing IPTV video quality monitoring system and the IPTV video quality monitoring system overall module schematic diagram that has used video quality assessment system of the present invention respectively;
  • Fig. 2 is a schematic diagram of network health index curve and stall curve trend fitting according to the present invention
  • Fig. 3 is a flow chart of a video quality assessment method based on a network health index according to the present invention.
  • each block in a flowchart or block diagram may represent a module, program segment, or part of code that includes one or more Executable instructions.
  • MDI is highly correlated and consistent with video quality. As an industry standard for testing the quality of IP video streaming, it can be used to measure the jitter and packet loss rate of network media streams. MDI contains two parameters: DF delay factor and MLR media packet loss rate.
  • the video frame information data in the chip buffer is the comprehensive reflection result of network transmission and set-top box processing. 500ms frame rate freeze and duration, the live network environment test shows that the effect is highly consistent with the human eye experience.
  • the present invention proposes an MDI-based network health index algorithm (including empirical formulas and judgment logic), and implants the algorithm into the client set-top box probe that cannot provide a chip freeze data interface, so that it can and has The client set-top box of the chip card data interface is also correctly evaluated for good and good rate.
  • an MDI-based network health index algorithm including empirical formulas and judgment logic
  • the acquisition of the MDI-based network health index algorithm requires optimization and correction of its algorithm coefficients in the laboratory in combination with network indicators and stall curve fitting.
  • FIG. 1 are respectively the overall module schematic diagram of the existing IPTV video quality monitoring system and the overall module schematic diagram of the IPTV video quality monitoring system using the video quality evaluation system of the present invention.
  • set-top box A is an existing set-top box with a chip frame freeze information statistics module, which is common in southern provinces. It can already realize the objective quality assessment based on the frame information analysis frame information provided by the chip without reference type, without the need of the present invention.
  • the sub-method is shown by a dotted line in the figure; and the set-top box B is an existing set-top box without a chip frame freeze information statistics module, as is common in northern provinces, and cannot implement frame information analysis based on the chip. There is no reference type for frame freeze data
  • the objective quality assessment of , the method of the present invention is proposed for such a set-top box B.
  • the set-top box B' among Fig. 1 (b) is to increase the network health index (h_mos) MDI algorithm module of the present invention in the existing set-top box B that does not have the chip frame stall information statistics module, so that it can realize the same as the set-top box A based on Video Quality Assessment for the Network Health Index.
  • h_mos network health index
  • the frame information statistics provided by the analysis chip and the video freezing of the user terminal are consistent with the corresponding effect of the human eye experience.
  • the test set-top box of the network health index algorithm module and the chip freeze information statistics module corrects the network health index according to the freeze curve. It is also necessary to insert a network damage meter connected to the computer between the test set-top box and the IPTV platform video to complete the trend fitting work of the network health index curve and the freezing curve.
  • Fig. 2 is a schematic diagram of network health index curve and stall curve trend fitting according to the present invention, in the figure:
  • L1 is the chip freeze curve, that is, the curve that reflects the relationship between the number of chip freezes and the measured packet loss value
  • L2 is the initial network health index (h_mos) curve, that is, the calculation curve of the initial coefficient of the model and the packet loss value;
  • L3 is the fitting target network health index (h_mos) curve, that is, the final target curve of optimal fitting. It can be seen in Figure 2 that L3 and L1 are horizontally flipped and symmetrical.
  • the empirical calculation formula of the calculation probe network health index (h_mos) of the present invention is as follows:
  • h_mos (5-(int(MDI/a)) ⁇ b-((float)lost/c)) ⁇ (pow(d, e ⁇ lost ⁇ f))
  • lost is packet loss
  • the unit in the algorithm is 0.01%
  • lost is 5
  • the actual packet loss is 0.05%.
  • the a, b, c, d, e, and f in the formula are dynamic coefficients, and their different values will make the calculated value of h_mos model change dynamically.
  • the goal is to make the initial curve L2 gradually change to the target curve L3 in the figure, that is, it is in a state of horizontal flip symmetry with L1.
  • the fitting optimization correction process is as follows:
  • the final optimal fit determines the model coefficients.
  • the critical threshold of the network health index corresponding to the excellent rate of 2 video freezes: h_mos 4.8.
  • h_mos (5.0-((float)lost/50)) ⁇ (pow(2.72,-0.000394 ⁇ lost ⁇ 50));
  • h_mos (4.9-((float)lost/50)) ⁇ (pow(2.72,-0.000394 ⁇ lost ⁇ 50));
  • h_mos (4.8-((float)lost/50)) ⁇ (pow(2.72,-0.000394 ⁇ lost ⁇ 50));
  • h_mos (4.7-((float)lost/50)) ⁇ (pow(2.72,-0.000394 ⁇ lost ⁇ 50));
  • h_mos (4.6-((float)lost/50)) ⁇ (pow(2.72,-0.000394 ⁇ lost ⁇ 50));
  • h_mos (4.5-((float)lost/50)) ⁇ (pow(2.72,-0.000394 ⁇ lost ⁇ 50));
  • h_mos (4.4-((float)lost/50)) ⁇ (pow(2.72,-0.000394 ⁇ lost ⁇ 50));
  • h_mos (4.3-((float)lost/50))*(pow(2.72,-0.000394*lost*50)).
  • Fig. 3 is a flowchart of the quality assessment method according to the present invention.
  • the video quality assessment based on the network health index is fitted with the video quality assessment based on the frame information provided by the chip to analyze the freeze data, and the network health index based on the evaluation method of the simulated analysis freeze data is obtained
  • the evaluation algorithm module and apply it to a set-top box that does not have the ability to analyze freeze data, and perform a flow chart of quality evaluation.
  • Steps S310-S350 are steps carried out in the laboratory.
  • test set-top box there are two video quality modules of h_mos and chip freeze (this is different from the actual use scene, and the set-top box with the ability to analyze the chip freeze in actual use does not need to install the evaluation based on the network health index of the present invention. algorithm module).
  • step S310 an MDI-based network health index algorithm is run once at a fixed time interval (for example, 5 minutes, which can be understood by those skilled in the art and can be adjusted according to specific circumstances), and a h_mos data value is calculated; at the same time
  • a fixed time interval for example, 5 minutes, which can be understood by those skilled in the art and can be adjusted according to specific circumstances
  • step S320 the chip frame information jam information data is obtained every second, and a data value of the jam times is counted every interval of the same fixed time (for example, 5 minutes);
  • step S330 according to the h_mos obtained by the test and the stuck data value, record the curves respectively related to the control of network damage and packet loss;
  • step S340 the freeze data curve is basically kept unchanged, and at the same time, the coefficients of the network health index algorithm are corrected so that the h_mos value curve is similar to the freeze curve (see the above description in conjunction with FIG. 2 ).
  • Step S340 may include the process of fitting data multiple times, and finally correcting each parameter in the network health index algorithm;
  • step 350 the case of freezing twice is determined as the critical threshold of the network health index, that is, the critical h_mos threshold. And obtain the ideal algorithm module of video quality assessment based on network health index.
  • steps S360-S380 are steps for using scenarios.
  • step 360 the ideal algorithm module obtained is applied to an existing set-top box, that is, a set-top box (set-top box B in Fig. 1 ) that does not provide a chip card data interface, making it a set-top box B'.
  • step 370 the set-top box implanted with the aforementioned ideal algorithm module reports the h_mos data to the video platform every fixed time interval (for example, 5 minutes, those skilled in the art can understand, and can be adjusted according to specific conditions).
  • the video quality rate is calculated by the platform. If the data reported by the set-top box within 5 minutes is less than the h_mos threshold, it is a poor-quality block; if the set-top box has a total of more than 5% of poor-quality blocks on a certain day, it is a poor-quality user.
  • the technical solution of the present invention effectively combines the objective quality assessment method based on the network health index with the data of another objective quality assessment method in which the number of chip freezes is highly consistent with the human eye experience.
  • the network health index algorithm proposed is based on a more reasonable MDI network index, which has a higher correlation with the video playback freeze reflected by the video frame information of the set-top box chip, and the combination fit is reasonable.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)

Abstract

本发明提出一种基于网络健康指数的视频质量评估方法和系统。通过将基于网络健康指数的视频质量评估与基于芯片提供的帧卡顿信息分析卡顿数据的视频质量评估进行拟合，获得模拟分析卡顿数据的评估方法的基于网络健康指数的评估算法模块，将其植入缺乏分析卡顿数据能力的机顶盒，实现客观正确且与人眼体验对应效果一致的视频质量评估，解决全网质量监测统一评价标准的问题。

Description

一种基于网络健康指数的视频质量评估方法与系统 技术领域
本发明涉及互联网领域，尤其涉及一种针对IPTV等实时流的基于网络健康指数的视频质量评估方法与系统。
背景技术
随着互联网和多媒体技术的蓬勃发展，在线观看视频已经成为日常生活中主要的娱乐方式之一，用户对于观看体验的要求随着网络带宽的增加和在线视频资源的日益丰富而不断提高。用户端视频质量是影响用户观看体验的关键因素，因为即便是质量较高的原始视频，也很有可能经过网络传输并受到机顶盒处理能力影响而发生失真。因此，为改善用户体验，对用户端的视频质量进行有效地评估具有重要意义。
人眼对视频的主要感觉在三个方面：流畅度(是否有可察觉的卡顿)、清晰度、和延时度。其中，流畅度又是视频质量体验指标中最重要的一项。
目前，对视频质量进行评估主要有主观质量评估和客观质量评估，也可以主客观结合。
主观质量评估通常需要在受限制的环境条件下进行，人员分组观看后按等级主观打分，直接利用人眼的体验结果，实施难度较大，且不便于对大量视频进行评估，且个人打分因主观因素而存在较大偏差。
主客观结合法将客观质量评估结合个人平均打分，拟合效果不好，且同样因个人打分的主观因素而存在较大偏差。
客观质量评估根据其采用的参考模型可进一步分为全参考、部分参考、无参考三类：
1、全参考：其方法是对比原始视频进行评估，这样达到了像素级，可全面准确评价，与人眼效果一致，但是因为其需要有原始视频作对比而 原始视频往往无法获取，而且消耗的带宽太多，可应用范围受到限制，适应性差，视频分辨率缩放之后不具有可比性；
2、部分参考：其方法是对比部分原始视频并施加运算，这样便于实施，相比于全参考也可以减小带宽的开销，但是评估值不一定与人眼体验效果一致，存在一定偏差；
3、无参考：其方法包括分析视频帧信息、分析网络传输数据、分析芯片卡顿数据。
分析视频帧信息优点是比较准确，但算法复杂且结果难以与人眼体验对应；
分析网络丢包延时等传输数据，考虑丢包率和平均连续丢包数对视频质量的影响，还考虑了丢包离散度所指示的丢包分布对视频质量的影响，该方法较为简便，但评估值也无法与人眼体验对应，偏差很大；
分析卡顿数据，通过分析芯片提供的帧信息统计用户终端的视频卡顿，真实地反映用户的视频体验，可以实现与人眼体验对应效果一致，但需要大量的机顶盒芯片和播放器的适配工作，在无法提供芯片卡顿数据接口即没有芯片帧信息的场景中无法实现。
由于对于视频质量评估方法和系统好坏的判断而言，评估结果与人眼的主观体验感受一致程度是最为重要的指标。基于此，综合前述各质量评估方法可见，仅有全参考类型和分析卡顿数据的无参考类型两种可以实现与人眼体验对应效果一致，但其中全参考类型又因原始视频不易获取而应用范围受到限制。因此，只有后者，即基于芯片提供的帧信息分析卡顿数据无参考类型的客观质量评估才能较好的评估视频质量。
目前已经建立的IPTV视频质量监测系统中，南方省份大多通过机顶盒探针，读取播放器输出的芯片底层视频卡顿帧信息数据，如图1(a)中的机顶盒A具有芯片帧卡顿信息统计模块，基于芯片提供的帧信息分析卡顿数据可较好实现视频质量的评估。但是北方省份的机顶盒，如图1(a)中的机顶盒B，由于播放器和硬件问题，没有硬件芯片帧卡顿信息统计模块没有卡顿数据，或者虽然硬件芯片有卡顿数据但是播放器软件不予向探针传输卡顿数据，因此无法实现上述基于芯片提供的帧信息分析卡顿数据 的无参考类型的客观质量评估。这样全网中的那些没有芯片帧卡顿信息统计模块的机顶盒将导致全网规模统计视频质量评估难以进行。
因此，亟须针对无法提供芯片卡顿数据接口的客户端机顶盒，即如图1(a)中的机顶盒B，提出一种高效且便于实施，同时客观正确且与人眼体验对应效果一致的视频质量评估方法，解决全网质量监测以及南北省份统一评价标准的问题。
发明内容
提供本发明内容以便以简化形式介绍将在以下详细描述中进一步描述的一些概念。本发明内容并不旨在标识出所要求保护的主题的关键特征或必要特征；也不旨在用于确定或限制所要求保护的主题的范围。
本发明提出了一种在机顶盒探针内植入网络健康指数算法，获取媒体传输系数MDI(Media Delivery Index)值，结合网络指标和卡顿曲线拟合优化修正其算法系数，以此较佳地模拟分析卡顿数据的评估方法，进行网络健康指数计算和评估。本发明的技术方案实施难度小、准确性高。不需要全网机顶盒适配复杂的芯片帧信息卡顿信息库，使得播放器不能提供芯片信息的场景也能够正确地进行视频优良率评估。
本发明的视频质量评估系统，包括：视频平台；具备芯片帧卡顿信息统计模块的机顶盒；以及不具备芯片帧卡顿信息统计模块但植入了基于网络健康指数h_mos的视频质量评估算法模块的机顶盒。其中，基于h_mos的视频质量评估算法模块通过将基于MDI计算h_mos的视频质量评估与基于芯片提供的帧卡顿信息分析卡顿数据的视频质量评估进行拟合而获得。
本发明的视频质量评估方法，包括以下步骤：在位于实验室的同时具备基于h_mos的算法模块和基于芯片卡顿的评估模块的测试机顶盒中将基于MDI计算h_mos的视频质量评估与基于芯片提供的帧卡顿信息分析卡顿数据的视频质量评估进行拟合，获得模拟分析卡顿数据的评估效果的基于h_mos的评估的理想算法模块；将理想算法模块植入缺乏分析卡顿数据能力的机顶盒中；机顶盒以固定时间间隔向视频平台上报一次h_mos， 出现一次上报的h_mos低于h_mos临界阈值的情况就记录一个质差块，全天内质差块数量累积高于一比例的机顶盒列为质差用户；基于全天内质差用户数量和当天播放过视频的用户总数，根据公式“规模统计视频优良率＝(1-质差用户数/播放过视频的用户总数)×100％”来计算全天规模统计视频优良率，获得整体视频质量评估结果。
在前述测试机顶盒中进行的拟合过程中，每隔固定时间运行一次基于MDI的h_mos算法，计算得到一个h_mos数据值，记录h_mos数据值与网络损伤丢包控制关系的曲线；每秒获取芯片帧信息卡顿信息数据，每隔相同固定时间统计出一个卡顿次数数据值，记录卡顿次数数据值与网络损伤丢包控制关系的曲线；将卡顿次数数据值与网络损伤丢包控制关系的曲线保持不变，通过修正h_mos算法系数，拟合h_mos数据值与网络损伤丢包控制关系的曲线以与卡顿次数数据值与网络损伤丢包控制关系的曲线相似；以及确定卡顿两次的情况作为h_mos临界阈值。
通过阅读下面的详细描述并参考相关联的附图，这些及其他特点和优点将变得显而易见。应该理解，前面的概括说明和下面的详细描述只是说明性的，不会对所要求保护的各方面形成限制。
附图说明
以下将通过参考附图中示出的具体实施例来对本发明进行更具体描述。
图1的(a)和(b)分别是现有IPTV视频质量监测系统整体模块示意图和使用了本发明的视频质量评估系统的IPTV视频质量监测系统整体模块示意图；
图2是根据本发明的网络健康指数曲线与卡顿曲线趋势拟合示意图；
图3是根据本发明的基于网络健康指数的视频质量评估方法的流程图。
附图中的流程图和框图显示了根据本申请的实施例的系统、方法可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或代码的一部分，所述模块、程序段或代码的 一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。
具体实施方式
以下将通过参考附图中示出的具体实施例来对本发明进行更具体描述。通过阅读下文具体实施方式的详细描述，本发明的各种优点和益处对于本领域普通技术人员将变得清楚明了。然而应当理解，可以以各种形式实现本发明而不应被这里阐述的各实施方式所限制。提供以下实施方式是为了能够更透彻地理解本发明。除非另有说明，本申请使用的技术术语或者科学术语应当为本申请所属领域技术人员所理解的通常意义。
发明人注意到，媒体传输质量是影响和衡量网络视频质量的主要因素，机顶盒处理能力对网络视频质量的影响次之。
MDI与视频质量高度相关一致，作为IP视频流传输质量测试的行业标准，可以被用来衡量网络媒体流抖动和丢包率。MDI包含两个参数：DF延迟因子和MLR媒体丢包率。而芯片缓存内视频帧信息数据，是网络传输和机顶盒处理综合反映结果。500ms帧率卡顿和时长，经现网环境测试表明与人眼体验感觉效果高度一致。
基于此，本发明提出了一种基于MDI的网络健康指数算法(包括经验公式和判断逻辑)，将该算法植入无法提供芯片卡顿数据接口的客户端机顶盒探针内，使得其能够和具有芯片卡顿数据接口的客户端机顶盒一样正确地进行优良率评估。
而基于MDI的网络健康指数算法的获得，需要在实验室结合网络指标和卡顿曲线拟合优化修正其算法系数。
图1的(a)和(b)分别是现有IPTV视频质量监测系统整体模块示意图和使用了本发明的视频质量评估系统的IPTV视频质量监测系统整体模块示意图。
其中的机顶盒A为现有的具有芯片帧卡顿信息统计模块的机顶盒，如南方省份常见的，已经可以实现基于芯片提供的帧信息分析卡顿数据无参考类型的客观质量评估，无需本发明的分方法，在图中用虚线示出；而机顶盒B为现有的没有芯片帧卡顿信息统计模块的机顶盒，如北方省份常 见的，不能实现基于芯片提供的帧信息分析卡顿数据无参考类型的客观质量评估，本发明的方法是针对这样的机顶盒B而提出的。
图1(b)中的机顶盒B’为在现有的没有芯片帧卡顿信息统计模块的机顶盒B中增加本发明的网络健康指数(h_mos)MDI算法模块，使其能够和机顶盒A一样实现基于网络健康指数的视频质量评估。
为了使得基于网络健康指数的视频质量评估能够真实地反映用户的视频体验，像分析芯片提供的帧信息统计用户终端的视频卡顿实现与人眼体验对应效果一致，发明人在实验室采用同时具备网络健康指数算法模块和芯片卡顿信息统计模块的测试机顶盒，根据卡顿曲线对网络健康指数进行修正。还需要在测试机顶盒和IPTV平台视频之间插入连接到电脑的网络损伤仪来完成网络健康指数曲线与卡顿曲线趋势拟合工作。
图2是根据本发明的网络健康指数曲线与卡顿曲线趋势拟合示意图，图中：
L1为芯片卡顿曲线，即反映芯片卡顿次数与丢包值实测关系的曲线；
L2为初始网络健康指数(h_mos)曲线，即模型初始系数与丢包值计算曲线；
L3为拟合目标网络健康指数(h_mos)曲线，即优化拟合的最终目标曲线，在图2中可见L3与L1呈水平翻转对称状态。
本发明的计算探针网络健康指数(h_mos)的经验计算公式如下：
h_mos＝(5-(int(MDI/a))×b-((float)lost/c))×(pow(d，e×lost×f))
其中，lost为丢包，算法中单位为0.01％，lost为5，实际丢包为0.05％。
公式中的a、b、c、d、e、f为动态系数，它们的不同取值将使h_mos模型计算值动态变化。目标是使得图中初始曲线L2逐步变化到目标曲线L3，即与L1呈水平翻转对称状态。
拟合优化修正过程如下：
h_mos曲线函数，先在y＝2.5水平直线上横轴反转(y取2.5是因为：h_mos的y值域是0～5分，取中间值2.5翻转后，即可与卡顿数据曲线的函数单调性保持一致，方能拟合)
与卡顿曲线标幺量纲同一处理
再使用单纯形优化算法、最小二乘法
最终优化拟合确定模型系数。
在一个实施例中，根据实验室测试拟合结果，卡顿曲线拟合后系数取值：a＝50，b＝0.1，c＝50，d＝2.72，e＝-0.000394，f＝50。根据上述公式，对应视频卡顿2次的优良率的网络健康指数临界阈值：h_mos＝4.8。
运算逻辑经验步骤如下：(a＞10)
if(MDI＜＝10)
h_mos＝(5.0-((float)lost/50))×(pow(2.72，-0.000394×lost×50))；
else if(MDI＜＝a＝50)
h_mos＝(4.9-((float)lost/50))×(pow(2.72，-0.000394×lost×50))；
else if(MDI＜＝2a＝100)
h_mos＝(4.8-((float)lost/50))×(pow(2.72，-0.000394×lost×50))；
else if(MDI＜＝3a＝150)
h_mos＝(4.7-((float)lost/50))×(pow(2.72，-0.000394×lost×50))；
else if(MDI＜＝4a＝200)
h_mos＝(4.6-((float)lost/50))×(pow(2.72，-0.000394×lost×50))；
else if(MDI＜＝5a＝250)
h_mos＝(4.5-((float)lost/50))×(pow(2.72，-0.000394×lost×50))；
else if(MDI＜＝6a＝300)
h_mos＝(4.4-((float)lost/50))×(pow(2.72，-0.000394×lost×50))；
else
h_mos＝(4.3-((float)lost/50))×(pow(2.72，-0.000394×lost×50))。
图3是根据本发明的质量评估方法的流程图。
在实验室的测试机顶盒中，将基于网络健康指数的视频质量评估与基于芯片提供的帧信息分析卡顿数据的视频质量评估进行拟合，获得模拟分析卡顿数据的评估方法的基于网络健康指数的评估算法模块，并将其应用于没有分析卡顿数据能力的机顶盒，进行质量评估的流程图。
步骤S310-S350为实验室中进行的步骤。
在测试机顶盒中同时具备h_mos和芯片卡顿两种视频质量模块(这不同于实际使用场景，实际使用中具备分析芯片卡顿进行评估的能力的机顶盒无需加装本发明的基于网络健康指数的评估算法模块)。
在步骤S310，每间隔固定时间(例如5分钟，本领域技术人员可以理解，可根据具体情况进行调整)运行一次基于MDI的网络健康指数算法，计算出一个h_mos数据值；同时
在步骤S320，每秒获取芯片帧信息卡顿信息数据，每间隔相同固定时间(例如5分钟)统计出一个卡顿次数数据值；
在步骤S330，根据测试得到的h_mos和卡顿数据值，记录各自与网络损伤丢包控制关系的曲线；
在步骤S340，将卡顿数据曲线基本保持不变，同时修正网络健康指数算法系数，使h_mos数值曲线，与卡顿曲线趋势相似(参见上述结合图2的描述)。步骤S340可以包括多次拟合数据的过程，最终修正网络健康指数算法中各个参数；
在步骤350，确定卡顿2次的情况作为网络健康指数临界阈值，即临界h_mos阈值。并获得基于网络健康指数的视频质量评估的理想算法模块。
接下来，步骤S360-S380为使用场景的步骤。
在步骤360，将所获得的理想算法模块应用到现有机顶盒，即未提供芯片卡顿数据接口的机顶盒(图1中的机顶盒B)，使之成为机顶盒B’。
在步骤370，植入前述理想算法模块的机顶盒每间隔固定时间(例如5分钟，本领域技术人员可以理解，可根据具体情况进行调整)向视频平台上报一次h_mos数据。由平台计算视频优良率。如果机顶盒某个5分钟上报数据的小于h_mos阈值，则为一个质差块；如果机顶盒某天共计有高于5％质差块，则为质差用户。
在步骤380，计算一整天的规模统计视频优良率＝(1-质差用户数/播放过视频的用户总数)×100％。至此，获得整体视频质量评估结果。
本发明的技术方案使基于网络健康指数的客观质量评估方法与另外 一种人眼体验高一致性的芯片卡顿次数的客观质量评估方法的数据，有效进行结合。提出的网络健康指数算法，基于了更合理的MDI网络指标，其与机顶盒芯片视频帧信息反映的视频播放卡顿的相关性更高，结合拟合具有合理性。
以上各实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述各实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分或者全部技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的范围，其均应涵盖在本申请的权利要求和说明书的范围当中。

Claims (10)

  1. 一种视频质量评估方法，所述方法包括：
    在测试机顶盒中将基于媒体传输系数MDI计算网络健康指数h_mos的视频质量评估与基于芯片提供的帧卡顿信息分析卡顿数据的视频质量评估进行拟合，获得模拟分析卡顿数据的评估效果的基于网络健康指数的评估的理想算法模块；
    将所述理想算法模块植入缺乏分析卡顿数据能力的机顶盒中。
  2. 如权利要求1所述的方法，其特征在于，进一步包括：
    所述机顶盒以固定时间间隔向视频平台上报一次网络健康指数，出现一次上报的所述网络健康指数低于网络健康指数临界阈值的情况就记录一个质差块，全天内质差块数量累积高于一比例的机顶盒列为质差用户。
  3. 如权利要求1所述的方法，其特征在于，进一步包括：
    基于全天内质差用户数量和当天播放过视频的用户总数，计算全天规模统计视频优良率，获得整体视频质量评估结果。
  4. 如权利要求4所述的方法，其特征在于，所述规模统计视频优良率通过以下公式计算：
    规模统计视频优良率＝(1-质差用户数/播放过视频的用户总数)×100％。
  5. 如权利要求1所述的方法，其特征在于，所述测试机顶盒位于实验室，且同时具备基于网络健康指数的算法模块和基于芯片卡顿的评估模块，获得所述理想算法模块的步骤包括，在所述测试机顶盒中执行以下步骤：
    每隔固定时间运行一次基于媒体传输系数的网络健康指数算法，计算得到一个网络健康指数数据值，记录所述网络健康指数数据值与网络损伤丢包控制关系的曲线；
    每秒获取芯片帧信息卡顿信息数据，每隔相同固定时间统计出一个卡顿次数数据值，记录所述卡顿次数数据值与网络损伤丢包控制关系的曲线；
    将所述卡顿次数数据值与网络损伤丢包控制关系的曲线保持不变，通过修正网络健康指数算法系数，拟合所述网络健康指数数据值与网络损伤丢包控制关系的曲线以与所述卡顿次数数据值与网络损伤丢包控制关系的曲线相似；以及
    确定卡顿两次的情况作为网络健康指数临界阈值。
  6. 如权利要求5所述的方法，其特征在于，计算所述网络健康指数基于以下经验计算公式：
    h_mos＝(5-(int(MDI/a))×b-((float)lost/c))×(pow(d，e×lost×f))
    其中lost为丢包，a、b、c、d、e、f为动态系数。
  7. 如权利要求5所述的方法，其特征在于，所述拟合的步骤包括：
    将所述网络健康指数数据值与网络损伤丢包控制关系的曲线在y＝2.5的水平直线上横轴反转；
    与所述卡顿次数数据值与网络损伤丢包控制关系的曲线标幺量纲同一处理；以及
    使用单纯形优化算法、最小二乘法，优化拟合确定模型系数。
  8. 如权利要求5所述的方法，其特征在于，所述拟合后各系数取值：a＝50，b＝0.1，c＝50，d＝2.72，e＝-0.000394，f＝50。
  9. 一种视频质量评估系统，包括：
    视频平台；
    具备芯片帧卡顿信息统计模块的机顶盒；以及
    不具备芯片帧卡顿信息统计模块但植入了基于网络健康指数h_mos的视频质量评估算法模块的机顶盒。
  10. 如权利要求9所述的系统，其特征在于所述基于网络健康指数的视频质量评估算法模块通过将基于媒体传输系数MDI计算网络健康指数的视频质量评估与基于芯片提供的帧卡顿信息分析卡顿数据的视频质量评估进行拟合而获得。
PCT/CN2021/135674 2021-08-24 2021-12-06 一种基于网络健康指数的视频质量评估方法与系统 WO2023024316A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110974566.4A CN113852801A (zh) 2021-08-24 2021-08-24 一种基于网络健康指数的视频质量评估方法与系统
CN202110974566.4 2021-08-24

Publications (1)

Publication Number Publication Date
WO2023024316A1 true WO2023024316A1 (zh) 2023-03-02

Family

ID=78976093

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/135674 WO2023024316A1 (zh) 2021-08-24 2021-12-06 一种基于网络健康指数的视频质量评估方法与系统

Country Status (2)

Country Link
CN (1) CN113852801A (zh)
WO (1) WO2023024316A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117076281A (zh) * 2023-10-13 2023-11-17 晨达(广州)网络科技有限公司 一种基于深度学习的软件质量评估方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103596068A (zh) * 2013-11-01 2014-02-19 李常春 基于媒体丢包率指标的iptv业务健康度评价方法
US10574980B1 (en) * 2019-02-13 2020-02-25 Case On It, S.L. Determination of metrics describing quality of video data by extracting characteristics from the video data when communicated from a video source to a display device
CN111083125A (zh) * 2019-12-02 2020-04-28 上海交通大学 神经网络优化的无参考自适应流媒体质量评价方法及系统
CN111083465A (zh) * 2018-10-22 2020-04-28 中国电信股份有限公司 视频卡顿分析方法、装置和系统、用户终端和存储介质
CN112672143A (zh) * 2020-12-21 2021-04-16 北京金山云网络技术有限公司 视频质量的评估方法、装置及服务器

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109921941B (zh) * 2019-03-18 2021-09-17 腾讯科技（深圳）有限公司 网络业务质量评估和优化方法、装置、介质及电子设备
CN110049373B (zh) * 2019-04-29 2021-04-06 宜通世纪科技股份有限公司 机顶盒卡顿检测方法、系统及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103596068A (zh) * 2013-11-01 2014-02-19 李常春 基于媒体丢包率指标的iptv业务健康度评价方法
CN111083465A (zh) * 2018-10-22 2020-04-28 中国电信股份有限公司 视频卡顿分析方法、装置和系统、用户终端和存储介质
US10574980B1 (en) * 2019-02-13 2020-02-25 Case On It, S.L. Determination of metrics describing quality of video data by extracting characteristics from the video data when communicated from a video source to a display device
CN111083125A (zh) * 2019-12-02 2020-04-28 上海交通大学 神经网络优化的无参考自适应流媒体质量评价方法及系统
CN112672143A (zh) * 2020-12-21 2021-04-16 北京金山云网络技术有限公司 视频质量的评估方法、装置及服务器

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117076281A (zh) * 2023-10-13 2023-11-17 晨达(广州)网络科技有限公司 一种基于深度学习的软件质量评估方法

Also Published As

Publication number Publication date
CN113852801A (zh) 2021-12-28

Similar Documents

Publication Publication Date Title
US8405773B2 (en) Video communication quality estimation apparatus, method, and program
KR101464456B1 (ko) 비디오 데이터 품질 평가 방법 및 장치
Zadtootaghaj et al. Modeling gaming qoe: Towards the impact of frame rate and bit rate on cloud gaming
CN104796443B (zh) 一种移动流媒体用户体验质量QoE修正方法和服务器
Li et al. Modeling QoE of virtual reality video transmission over wireless networks
WO2023024316A1 (zh) 一种基于网络健康指数的视频质量评估方法与系统
CN107027023A (zh) 基于神经网络的VoIP无参考视频通信质量客观评价方法
US10541894B2 (en) Method for assessing the perceived quality of adaptive video streaming
CN101959063A (zh) 媒体内容传送的网络传输效应的评估系统和方法
Xu et al. On the properties of mean opinion scores for quality of experience management
Xue et al. Mobile video perception: New insights and adaptation strategies
CN101616315A (zh) 一种视频质量评价方法、装置和系统
CN108271016B (zh) 视频质量评估方法及装置
Singla et al. Assessing media qoe, simulator sickness and presence for omnidirectional videos with different test protocols
Mozhaeva et al. Constant subjective quality database: the research and device of generating video sequences of constant quality
CN117061791B (zh) 云视频帧自适应协作渲染方法、装置及计算机设备
CN103596068B (zh) 基于媒体丢包率指标的iptv业务健康度评价方法
CN111355949B (zh) 音视频多媒体数据库的构建及多媒体主观质量评价方法
CN117599412A (zh) 一种基于云游戏业务质量检测的自适应渲染系统及方法
CN101895787B (zh) 一种视频编码性能主观评价方法及系统
CN111277899A (zh) 基于短期记忆和用户期望的视频质量评价方法
Wang et al. A video quality assessment method using subjective and objective mapping stategy
Liu et al. A quality-of-experience database for adaptive omnidirectional video streaming
Sun et al. Quality-of-Experience Assessment for Ultra-Low Latency Live Streaming Videos
Mozhaeva et al. NRspttemVQA: Real-Time Video Quality Assessment Based on the User’s Visual Perception

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21954832

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE