Analysis and prediction of QoE for video streaming using empirical measurements Funded by Forthnet, the GSRT with a Research Excellence grant, and by a.

Slides:



Advertisements
Similar presentations
1 P2P Layered Streaming for Heterogeneous Networks in PPSP K. Wu, Z. Lei, D. Chiu James Zhibin Lei 17/03/2010.
Advertisements

Lookback Scheduling for Long-Term Quality-of-Service Over Multiple Cells Hatem Abou-zeid*, Stefan Valentin, Hossam S. Hassanein*, and Mohamed F. Feteiha.
1 Developing a Predictive Model for Internet Video Quality-of-Experience Athula Balachandran, Vyas Sekar, Aditya Akella, Srinivasan Seshan, Ion Stoica,
LYU9903 QoS Schemes in Wireless Networks Ho Pun Mo & Ng Maggie Supervised by Prof. Lyu R.T. Michael 6 December, 1999.
Playback-buffer Equalization For Streaming Media Using Stateless Transport Prioritization By Wai-tian Tan, Weidong Cui and John G. Apostolopoulos Presented.
Doc.: IEEE /0604r1 Submission May 2014 Slide 1 Modeling and Evaluating Variable Bit rate Video Steaming for ax Date: Authors:
ITU Regional Standardization Forum For Africa Dakar, Senegal, March 2015 QoS/QoE Assessment Methodologies (Subjective and Objective Evaluation Methods)
Accurate & scalable models for wireless traffic workload Assistant Professor Department of Computer Science, University of Crete & Institute of Computer.
CStream: Neighborhood Bandwidth Aggregation For Better Video Streaming Thangam Vedagiri Seenivasan Advisor: Mark Claypool Reader: Robert Kinicki 1 M.S.
1 Voice Over IP (VoIP) is a common technology for performing voice calls over the Internet. Voice Over IP (VoIP) is a common technology for performing.
CROSS LAYER DESIGN CMPT 820: Multimedia Systems Kaushik Choudhary.
A Quality-Driven Decision Engine for Live Video Transmission under Service-Oriented Architecture DALEI WU, SONG CI, HAIYAN LUO, UNIVERSITY OF NEBRASKA-LINCOLN.
Performance Evaluation of IP Telephony over University Network A project funded by University Fast Track By M. Kousa, M Sait, A. Shafi, A. Khan King Fahd.
Quality of Service in IN-home digital networks Alina Albu 23 October 2003.
Designing QoE experiments to evaluate Peer-to-Peer streaming applications Tom Z.J. Fu, CUHK Dah Ming Chiu, CUHK Zhibin Lei, ASTRI VCIP 2010, Huang Shan,
Available bandwidth measurement as simple as running wget D. Antoniades, M. Athanatos, A. Papadogiannakis, P. Markatos Institute of Computer Science (ICS),
CS541 Advanced Networking 1 Cognitive Radio Networks Neil Tang 1/28/2009.
UNC/FORTH Archive of Wireless Traces, Models and Tools 1 Foundation for Research & Technology-Hellas (FORTH) & University of Crete 2 University of North.
1 Prof. Maria Papadopouli University of Crete ICS-FORTH Lecture on VoIP over wireless networks.
Nov. 3, 2000 Adaptive Playout Scheduling in Packet Voice Communications.
Q.Int_speed_test Draft Recommendation Eva Ibarrola SG11 Q15 Acting Rapporteur JCA-CIT meeting (electronic meeting)
Enhancing TCP Fairness in Ad Hoc Wireless Networks Using Neighborhood RED Kaixin Xu, Mario Gerla University of California, Los Angeles {xkx,
The Effects of Latency on Player Performance in Cloud-based Games Mark Claypool and David Finkel Computer Science and Interactive.
Ch. 28 Q and A IS 333 Spring Q1 Q: What is network latency? 1.Changes in delay and duration of the changes 2.time required to transfer data across.
Statistical Natural Language Processing. What is NLP?  Natural Language Processing (NLP), or Computational Linguistics, is concerned with theoretical.
Doc.: IEEE /1159r1 Submission Sept 2013 Guoqing Li (Intel)Slide 1 Video Performance Requirements and Simulation Parameters Date: Authors:
Doc.: IEEE /0787r0 Submission July 2013 Wu TianyuSlide 1 Follow-up Discussions on HEW Functional Requirements Date: Authors:
COGNITIVE RADIO FOR NEXT-GENERATION WIRELESS NETWORKS: AN APPROACH TO OPPORTUNISTIC CHANNEL SELECTION IN IEEE BASED WIRELESS MESH Dusit Niyato,
Slide Image Retrieval: A Preliminary Study Guo Min Liew and Min-Yen Kan National University of Singapore Web IR / NLP Group (WING)
Machine Learning1 Machine Learning: Summary Greg Grudic CSCI-4830.
1 Mobility Aware Server Selection for Mobile Streaming Multimedia CDN Muhammad Mukarram Bin Tariq, Ravi Jain, Toshiro Kawahara {tariq, jain,
Intelligent and Adaptive Middleware to Improve User-Perceived QoS in Multimedia Applications Pedro M. Ruiz, Juan A. Botia, Antonio Gomez-Skarmeta University.
Analysis of the Impact and Interactions of Protocol and Environmental Parameters on Overall MANET Performance Michael W. Totaro and Dmitri D. Perkins Center.
Wireless Networks Breakout Session Summary September 21, 2012.
RecSys 2011 Review Qi Zhao Outline Overview Sessions – Algorithms – Recommenders and the Social Web – Multi-dimensional Recommendation, Context-
1 Requirements for the Transmission of Streaming Video in Mobile Wireless Networks Vasos Vassiliou, Pavlos Antoniou, Iraklis Giannakou, and Andreas Pitsillides.
Chapter 5: Summary r principles behind data link layer services: m error detection, correction m multiple access protocols m link layer addressing, ARP.
Colombia, September 2013 The importance of models and procedures for planning, monitoring and control in the provision of communications services.
Computer Networks Performance Metrics. Performance Metrics Outline Generic Performance Metrics Network performance Measures Components of Hop and End-to-End.
Content Clustering Based Video Quality Prediction Model for MPEG4 Video Streaming over Wireless Networks Asiya Khan, Lingfen Sun & Emmanuel Ifeachor 16.
Fen Hou and Pin-Han Ho Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Ontario Wireless Communications and Mobile.
Delivering Adaptive Scalable Video over the Wireless Internet Pavlos Antoniou, Vasos Vassiliou and Andreas Pitsillides Computer Science Department University.
Quality of Service Karrie Karahalios Spring 2007.
Spatio-Temporal Modeling of Traffic Workload in a Campus WLAN Felix Hernandez-Campos 3 Merkouris Karaliopoulos 2 Maria Papadopouli 1,2,3 Haipeng Shen 2.
1 P2P Layer Streaming for Heterogeneous Networks in PPSP K. Wu, Z. Lei, D. Chiu Kent Kangheng Wu 9/11/2010.
Network Instruments VoIP Analysis. VoIP Basics  What is VoIP?  Packetized voice traffic sent over an IP network  Competes with other traffic on the.
Fair Class-Based Downlink Scheduling with Revenue Considerations in Next Generation Broadband wireless Access Systems Bader Al-Manthari, Member, IEEE,
New Models for Perceived Voice Quality Prediction and their Applications in Playout Buffer Optimization for VoIP Networks University of Plymouth United.
1 Presented by Jari Korhonen Centre for Quantifiable Quality of Service in Communication Systems (Q2S) Norwegian University of Science and Technology (NTNU)
University of Plymouth United Kingdom {L.Sun; ICC 2002, New York, USA1 Lingfen Sun Emmanuel Ifeachor Perceived Speech Quality.
Exploiting Context Analysis for Combining Multiple Entity Resolution Systems -Ramu Bandaru Zhaoqi Chen Dmitri V.kalashnikov Sharad Mehrotra.
1 Network Emulation Mihai Ivanovici Dr. Razvan Beuran Dr. Neil Davies.
Wireless Trace Analysis. Project Goals Summary of project goals: First goal: analyze wireless access patterns Second goal: implement Markov predictor.
Doc.: IEEE /xxxx Submission July 2007 Lei Du, DoCoMo Beijing Labs Slide 1 End-to-End QoS awareness for admission control Date: Authors:
Kyocera Corporation Enhance radio network connectivity and maintain a Quality of IP service application Proposal of extension of IEEE /7/16.
ITU Workshop on “Voice and Video over LTE” Geneva, Switzerland, 1 December 2015 Considerations for end to end video quality QoE assessment as a means of.
AIMS’99 Workshop Heidelberg, May 1999 Assessing Audio Visual Quality P905 - AQUAVIT Assessment of Quality for audio-visual signals over Internet.
Understanding the Impact of Network Dynamics on Mobile Video User Engagement M. Zubair Shafiq (Michigan State University) Jeffrey Erman (AT&T Labs - Research)
Modeling Web Quality-of-Experience on Cellular Networks
Web: ~ laoudias/pages/platform.htmlhttp://www2.ucy.ac.cy/ ~ laoudias/pages/platform.html
Development of a QoE Model Himadeepa Karlapudi 03/07/03.
1 Video and Voice over IP performance over a Satellite link Bob Dixon, Ohio State University/OARnet Prasad Calyam, OARnet Joint Techs Workshops, Columbus,
Saving Bitrate vs. Users: Where is the Break-Even Point in Mobile Video Quality? ACM MM’11 Presenter: Piggy Date:
Performance Comparison of Ad Hoc Network Routing Protocols Presented by Venkata Suresh Tamminiedi Computer Science Department Georgia State University.
Teng Wei and Xinyu Zhang
Vengatanathan Krishnamoorthi, Niklas Carlsson
Video through a Crystal Ball:
Muhammad Niswar Graduate School of Information Science
Modeling and Evaluating Variable Bit rate Video Steaming for ax
Presentation transcript:

Analysis and prediction of QoE for video streaming using empirical measurements Funded by Forthnet, the GSRT with a Research Excellence grant, and by a Google Faculty Award, 2013 (PI Maria Papadopouli) University of Crete Foundation for Research & Technology – Hellas (FORTH)

2 Users Objective measurements Using various tools: SNMP, syslog, wireshark, spectrum analyzers, Appscope, DAG cards Subjective measurements Using questionnaires performing field studies 2 Monitoring of wireless network environments

Network QoS metrics: throughput, jitter, round-trip delay, startup delay, packet loss, traffic load Application metrics: rebuffering events, video resolution, adaptation of video streaming

Definitions of QoE Overall acceptability of an application or service, as perceived subjectively by the end user includes the complete end-to-end system effects, and may be influenced by user expectations and context [ITU]. The degree of delight or annoyance of a person whose experiencing involves an application, service, or system. It results from the person's evaluation of the fulfillment of his or her expectations and needs with respect to the utility and/or enjoyment in the light of the person's context, personality and current state [Raake].

6 QoE metrics are characterized with techno-socio-economic-psychological terms QoS network metrics and application-based metrics Preference on QoS or price Price, willingness to pay Perceived QoE (e.g., opinion score) Intrinsic indicators towards a service provider e.g., its brand name, perceived value/reliability content (size, searching mechanisms) To define user experience is a very hard problem & to monetize it even harder!

7

Modeling approaches

9 Examples of QoS metrics for telecom services Achievable data rate Throughput, delay, packet loss Number of resource units TDMA: time slots Weber-Fechner Law IQX hypothesis Customer Satisfaction QoE metrics with techno-socio-economic-psychological terms Preference on QoS or price Price, willingness to pay Perceived QoE (e.g., opinion score) Intrinsic indicators towards a service provider e.g., its brand name, perceived value/reliability content (size, searching mechanisms) To define user experience is a very hard problem & to monetize it even harder!

Mathematical models of QoE Weber-Fechner Law IQX OoS: could be a network metric like average data rate, packet loss, delay

12 Our approach for predicting the QoE: Develop user-centric, service-oriented models based on network metrics Apply machine learning and data mining algorithms, such as: Decision Trees, Support Vector Regression, Artificial Neural Networks, Gaussian Naïve Bayes Find the set of predictors that minimizes the mean absolute error of a model (feature selection) Train the models based on empirical measurements collected from field studies We have demonstrated this methodology for VoIP, audio & video streaming

MLQoE: QoE prediction based on machine learning (ML) algorithms Takes as input the training set of the performance estimation loop, cross-validates it, and reports the best model dynamically. Estimates the performance of the best model in each fold and reports (as output) the mean error for the dataset. On user-centric modular QoE prediction for VoIP based on machine-learning alg. [IEEE Trans. on Mobile Computing]

NovaGO HF Tracker: Παράδειγμα Γραφικού Περιβάλλοντος σε έναν Χρήστη

System Architecture

Example of user activity Server GUI)

Field study summary Duration: 56 days (29 July—12 September) Clients: users never uploaded video sessions 20 users with at least 1 labeled video session (sessions rated with QoE score) 13 users with more than 5 labeled video sessions (considered for QoE prediction) Video sessions: 298 from 20 users: visual exploration 281 from 13 users: QoE prediction

298 video sessions – 20 users Video sessions per user QoE histogram 7 users (17 sessions) excluded from QoE prediction Few sessions with low QoE

Video session features 105 features extracted about the video sessions: Service type, startup delay, session duration, termination type, QoE score Buffering events number, {total, min, max, …} duration Video resolution (per-resolution) number of switches, {min, max, mean, …} resolution Network performance mean packet loss ratio, mean jitter, {min, max, …} signal strength User activity (pause, seek, off-screen events) number, {total, min, max, …} duration Respective statistics for the last 15, 30, 60 seconds of the session

Stationary Vs. Roaming Stationary293 (98.32%) Roaming5 (1.68%) Roaming video sessions are rare Stationary sessions: Smartphone associated with only 1 AP during the entire session Roaming sessions: Smartphone performed handovers between APs during the session

Directly user-perceived parameters Per-session statistics about the video playback 22 sessions with startup delay > 10 sec 8 video sessions that never started playing (startup delay = session duration) 20 sessions with buffering event duration ratio > 0.1 Weighted mean x-axis video resolution

Network conditions Per-session statistics about the network performance 5 video sessions with packet loss > 20%

Users perceive the degradation (low QoE scores) for startup delay >= 10 sec Related research reported that a startup delay beyond 2 sec causes viewers to abandon the video [Krishnan 13] Our speculation: Smartphone users are more tolerant

The higher the buffering ratio, the smaller the duration of the session Increase in buffering ratio can decrease viewing time [Krishnan 13, Dobrian 11]

Improving the data rate adaptation could reduce the buffering ratio Sessions of high resolution, poor connectivity high buffering duration ratio & low QoE scores Decreasing resolution might improve QoE

Sessions with higher startup delay, buffering ratio & lower network performance have lower QoE

Poor network performance during the last 15 sec of the session results in termination due to poor connectivity

Interesting Sessions in the Forthnet dataset Sessions with high buffering ratio, high duration and high score Sessions with poor connectivity status, rated with high QoE scores Sessions with degraded performance, rated with high QoE scores Their presence motivate us to perform a second (more controlled) field study

Different types of field studies for data collection Tradeoffs between: small-scale studies with homogeneous settings in non-controlled environments vs. larger-scale (potentially crowd-sensing/sourcing participatory) studies that can reach more people, representing a more realistic set of scenarios/conditions but with several unknown difficult to control dynamic exogenous parameters and heterogeneous settings. Challenges: Obtaining reliable measurements in such crowd-sourcing/sensing non-controlled field studies In general, it is difficult to obtain the “ground truth” about the QoE. The above also highlight the tension between subjectivity and reliability in the collected data.

FORTH dataset 50 produced videos 20 participants each user viewed all 50 videos duration 13 days 4 different reference videos (high quality) 4 chunks per video (duration 5 sec each one)

Playback video parameterized based on: Startup delay Number of buffering events Ratio of buffering duration Times when buffering events occur Duration of each buffering event Video resolutions for each chunk Aggregate resolution of the video

Subjectivity of the assessments & user sensitivity to different types of impairment Three scenarios: large startup delay number of rebuffering events low resolution Depending on the type of impairment appeared: some users are more tolerant/strict than others some users are more tolerant to some types of impairment & more strict to others statistically significance difference of the scores of users for the various types of impairment (T-student test)

Parameters with dominant impact on the QoE Forthnet dataset Aggregate approach: termination type of the session buffering events frequency weighted mean video resolution ratio packet loss User-centric approach: termination type (10 users) mean jitter for (6 users) startup delay and its ratio (5 users) packet loss for (4 users) the weighted mean video resolution and its ratio (4 users)

User-centric model: number of buffering events buffering ratio consistently Parameters with Dominant Impact on the QoE FORTH dataset

QoE prediction Forthnet dataset AlgorithmMeanMedianStd uQoE aggregate uQoE user-centric Users with high prediction error: Users with less than 15 sessions Users have pathologies (sessions with high degradations & high QoE) QoE scores are not evenly distributed

Slide 39 QoE prediction FORTH dataset AlgorithmMeanMedianStd uQoE user-centric