Accomplishing Seeing Time Versatility in Portable Video Spilling Utilizing Adaptable Video Coding.

Uploaded on:
[Theorem] The rest period proportion of cell phones getting the substream at edge rate and quan. step is: We compose , where mW is the force utilization ...
Slide 1

Accomplishing Viewing Time Scalability in Mobile Video Streaming Using Scalable Video Coding Cheng-Hsin Hsu Senior Research Scientist Deutsche Telekom R&D Lab USA Los Altos, CA Joint Work with Mohamed Hefeeda February 22, 2010

Slide 2

Recent studies demonstrate that numerous clients watch recordings utilizing their cell phones notwithstanding when stationary gadgets, for example, TVs, are accessible [Vorbau et al ., HotMobile\'07] Mobile gadgets have strict force limitation Battery advancements do not develop as quick as CPU rate, memory size, and plate limit Mobile Video Streaming 2

Slide 3

Mobile clients must consider battery lifetme as another measurement of survey quality For instance: a client watches a 30-min scene Motivation The most noteworthy conceivable quality High quality all through the scene Until coming up short on battery Quality Time 0 30 25 Streaming at a lower quality, in any case, requires transcoding, which is computationally serious 3

Slide 4

Multiple substreams can be effectively extricated from a versatile stream for backing distinctive resolutions, outline rates, and loyalty levels Often utilized for gadgets with heterogeneous assets Such as transmission capacity, show determination, and decoder ability We highlight another advantage of SVC: to empower seeing time adaptability Scalable Video Coding (SVC) 4

Slide 5

Goal : permit clients to pick coveted video quality and review time of a gushing video We first build up a system to foresee battery lifetime and saw nature of individual substreams separated from an adaptable stream We then propose a calculation and give clients a control handle to exchange off saw quality and review time Problem Statement 5

Slide 6

Quality-Power Adaptation Framework Desired Quality (in MOS) Desired Viewing Time (in hr) Framework Perceived Quality Model Comm. Power Model Video Characteristics Optimal Substream Adaptation Algorithm Other Power Model CPU Power Model Device Characteristics Required CPU Cycles Expected Viewing Time (in hr) Expected Quality (in MOS) 6

Slide 7

Divide power utilization into CPU: for video unraveling, composed as p c ( y ), where y is number of cycles every second Comm.: for getting video stream, composed as p n ( ) , where is the rest time frame proportion Background: represents others, for example, show backdrop illumination, composed as a steady p b Power Model 7

Slide 8

Depends on sorts of remote systems Such as WLANs, WiMAX, and 3G Cellular We concentrate on versatile TV arranges and propose a communicate plan, FMVB , for adaptable streams FMVB ( Flexible Mobile Video Broadcast ) underpins substream extractions at portable collectors Extractions must be done at recipients as a result of the communicate nature Enables both worldly (determined by edge rate t ) and loyalty (indicated by quan. step ) adaptability Communication Power Model 8

Slide 9

FMVB decides time and size of transmission blasts for individual layers Formulas are in the paper [Theorem] FMVB produces doable plans with blasts that are sufficiently long to support smooth playout Communication Power Model (cont.) Ch. 1, Layer Ch. 1, Layer Ch. 1, Layer Ch. 2, Layer Ch. 2, Layer Ch. 1, Layer Ch. 2, Layer Ch. 2, Layer Time 9

Slide 10

[Theorem] The rest time frame proportion of cell phones accepting the substream at edge rate and quan. step is: We compose , where mW is the force utilization of the system chip Communication Power Model (cont.) Our Comm. Model 10

Slide 11

Complexity model: gauge , which is the quantity of cycles required by translating a substream at edge rate and quan. step We can evaluate what number of cycles required for each substream, and after that? CPU Power Model [Ma and Wang, ICASSP\'08]. 11

Slide 12

Modern CPUs can keep running at a lower recurrence to spare vitality We coordinate produce (CPU recurrence) with interest (number of cycles) We embrace the vitality scaling model: Combining and capacities gives our CPU power model CPU Power Model (cont.) [He et al ., CSVT\'05]. CPU power productivity component 12

Slide 13

Total Power Consumption Battery limit is frequently measured in mAh utilizing worked as a part of circuits Viewing time can be composed as , where is the battery voltage Total Power Consumption Total CPU Comm. Bg. Number of cycles every second Interface rest period proportion 13

Slide 14

Quality Model MOS Score 0% - 100% 75% Perceived Quality Model Full Quality 70% 45% The apparent quality in MOS is given as: [Wang et al ., PV\'09]. 14

Slide 15

Leverages on the models to process expected review time (or quality) for each substream Returns the best substream while fulfilling clients\' prerequisites Quality-Power Adaptation Algorithms Optimal Substream QPAA Desired Viewing Time Expected Quality Optimal Substream QPAA Desired Quality Expected Viewing Time 15

Slide 16

Energy Saving on Mobile TV Receiver Use off-the-rack equipment Build programming in the base station Transmits DVB-H agreeable signs [MM08\'Demo] 16

Slide 17

Four model parameters are gadget subordinate : system interface power utilization : system interface move time : foundation power utilization : CPU power productivity component We exhibit how to determine the parameters utilizing a Nokia N96 telephone Device Dependent Parameter Estimation 17

Slide 18

We encode a 10-min news into a stream at video rate 450 kbps and sound rate 32 kbps We arrange our base station to communicate this stream with various between burst periods: 250, 500, 1000, 2000, and 3000 msec We utilize a telephone to observe every communicate for 3.5 hours, and we measure power utilization 4 times every second We utilize the estimations to infer mW and msec Network Interface Power Consumption and Transition Time 18

Slide 19

Background Power Consumption and CPU Efficiency Factor To quantify foundation power utilization, we design a N96 telephone to show pictures in slide show mode for 1.5 hr Isolate comm. furthermore, CPU power utilization Average bg. power utilization is p b = 290.38 mW With all other model parameters, we can figure the CPU power utilization, and in this way normal CPU Efficiency Factor: 19

Slide 20

Tradeoff between Video Quality and Power Consumption Use a N96 telephone with 950 mAh, 3.7 V battery Configure base station to communicate at 8.289 Mbps Consider Crew grouping with edge rates 3.75, 7.5, 15, and 30 fps and quan. steps 16, 40, 64, and 104 Three communicate administrations utilizing the proposed FMVB plan Temporal just, quality just, and consolidated 20

Slide 21

Sample Results Wide scope of review time versatility is conceivable: 4.3 to 11.1 hrs 21

Slide 22

Sample Results (cont.) Quality corruption is not direct to survey time increment Frame rate 30  15 copies seeing time, yet just decreases quality by 5% Highlights the significance of our structure 22

Slide 23

Sample Results (cont.) Combined adaptability: every point speaks to a substream Our structure helps clients to pick the best substream 23

Slide 24

Conclusions Proposed a quality-power adjustment system to methodicallly control the tradeoff between video quality and survey time Proposed a video communicate plan (FMVB) to communicate versatile streams for survey time adaptability Presented the estimation of gadget ward parameters, and showed the viability of the proposed adjustment system Framework can be stretched out to different remote systems and applications 24

Slide 25

Questions and Comments 25

Slide 26

Network Interface Power Consumption and Transition Time (cont.) Longer between burst periods  less spikes CDFs demonstrate that the distinction between chip rest/dynamic modes is mW 26

Slide 27

Network Interface Power Consumption and Transition Time (cont.) Broadcast plans with littler between burst periods  more occurrences Comparing any two plans gives us a worth msec, which is adjusted to 150 msec provided details regarding a late DVB-H chip datasheet [Philips] 27

View more...