Page 147 - Proceedings of the 2017 ITU Kaleidoscope
P. 147

Challenges for a data-driven society






















                                                                           Fig. 8. Inter packet delay
                         Fig. 6. The VQM observation
                                                              6.7.  Experimental Original  and Decoded  Sequence
           6.5. Three-component  Structural  Similarity  (3-  of Frames
           SSIM)Measurement
                                                              Figure 9 shows the  few  original and decoded video
           3-SSIM is a form of SSIM that is calculated as a weighted   sequences captured during live streaming experiment at the
           average of  SSIM  for three categories of regions: edges,   server and client respectively. The display resolution is
           textures, and smooth regions. From Figure 7, it is evident   640*480  with the  frame rate of 24. The live  video
           that the proposed SBQA-SP approach shows a higher 3-  streaming was carried out in a class room environment.The
           SSIM value compared to other approaches.  The average   codec used for encoding and decoding process is H.264 for
           values  shows that SBQA-SP approach provides quality   live as well as stored video.
           which is 0.4% higher than  QBQA approach and 1.04%
           higher than SBQA-GP approach  with respect to 3-SSIM
           scale.



                                                              (a) Original






                                                              (b) Decoded

                                                                  Fig. 9. Live Streaming (few selectedoriginal and
                                                                               decoded frames)


                        Fig. 7. The 3-SSIMindex                     7. CONCLUSION AND FUTURE WORK
           6.6. Inter Arrival Packet Delay                    An HTTP adaptive streaming through 4G wireless network
                                                              was implemented using Q-learning with NR metrics for live
           Figure 8 shows theobserved  inter arrival packet delay on   video considering the challenges due to  varying  wireless
           the Airtel4G LTE-TD network during live video streaming.   link condition and internet traffic. To achieve  maximum
           An average delay of 7.6 millisecondswas observed during   obtainable quality, the decisions that are  made to adapt
           the experimentation process.                       video quality in a real time video streaming were based on
                                                              the current state of the system and the action to be taken in
                                                              that state that maximizes the reward. The proposed SBQA-
                                                              SP algorithm  using Softmax policy  was  formulated
                                                              andimplemented using ITU-T G.1070 model  for no-
                                                              reference metrics quality evaluation. Since ITU-T G.1070 is
                                                              targeted atQoE / QoS in  video telephony,  it constrains in
                                                              achieving  higher quality of streaming video in an IP
                                                              network.
                                                              The framework    was implemented in one      way
                                                              communication but it can easily be extended to support
                                                              two-way video communication. On the other aspect, the





                                                          – 131 –
   142   143   144   145   146   147   148   149   150   151   152