A time-varying subjective quality model for mobile streaming videos with stalling events
Computing & Wireless : Application Software
Available for licensing
- Alan Bovik, Ph.D. , Electrical and Computer Engineering
- Janice Pan , Electrical and Computer Engineering
- Phani Ghadiyaram
Streaming videos on mobile devices is becoming increasingly popular, but this practice is affected by startup delays and stalling events due to network impairments. These delays in streaming service can greatly impact a viewer’s quality of experience (QoE). Thus, having a tool that can automatically and accurately predict a viewer’s QoE, as related to start-up delays and rebuffering events, can help optimize network streaming services so that a user’s overall QoE is maximized.
Researchers at The University of Texas at Austin have developed a dynamic model that can predict subjective QoE on videos that have been afflicted by stalling events. This model is perceptually-based and thus models potential linearities and nonlinearities in the human visual system (HVS) in addition to hysteresis or the recency effect.
- Simply structured
- Computationally efficient
- Highly accurate
- Allows for flexibility in model design, since other QoE-influencing inputs such as bitrate, audio, spatial and temporal quality information can be added to the existing model at very little additional computational cost
Standard system identification tools are used to train the parameters of this model, using a number of novel input functions, which are descriptive of a video´s unique stalling pattern.
Video streaming companies such as Netflix, Amazon Video, Hulu, HBO, etc.
Beta product/commercial prototype
- 1 U.S. patent application filed