Modeling Continuous Video QoE Evolution: A State Space Approach

Abstract

A rapid increase in the video traffic together with an increasing demand for higher quality videos has put a significant load on content delivery networks in the recent years. Due to the relatively limited delivery infrastructure, the video users in HTTP streaming often encounter dynamically varying quality over time due to rate adaptation, while the delays in video packet arrivals result in rebuffering events. The user quality-of-experience (QoE) degrades and varies with time because of these factors. Thus, it is imperative to monitor the QoE continuously in order to minimize these degradations and deliver an optimized QoE to the users. Towards this end, we propose a nonlinear state space model for efficiently and effectively predicting the user QoE on a continuous time basis. The QoE prediction using the proposed approach relies on a state space that is defined by a set of carefully chosen time varying QoE determining features. An evaluation of the proposed approach conducted on two publicly available continuous QoE databases shows a superior QoE prediction performance over the state-of-the-art QoE modeling approaches. The evaluation results also demonstrate the efficacy of the selected features and the model order employed for predicting the QoE. Finally, we show that the proposed model is completely state controllable and observable, so that the potential of state space modeling approaches can be exploited for further improving QoE prediction.

Streaming Video QoE Modeling and Prediction: A Long Short-Term Memory Approach

“…3) Time elapsed since last rebuffering (T R ): Since, the user QoE is heavily influenced by the occurrence of rebuffering events, we employ T R , a variable to keep track of the time elapsed since the occurrence of the last rebuffering event. T R has been used for QoE prediction in [33]. We subsequently show that the proposed model driven by these features is powerful enough to provide superior prediction that significantly outperforms the state-of-the-art QoE prediction models.…”

“…Due to their demonstrated efficiency, we employ the following three features for QoE prediction in the proposed LSTM-QoE [5], [33]:…”

“…STSQ can be measured using offthe-shelf video quality assessment (VQA) metrics such as STRRED [23], MS-SSIM [30]. STSQ as a feature has been successfully employed for QoE prediction in [5] and [33].…”

“…The SSLs provide a state space interpretation for modeling the nonlinear non-Markovian dynamics of LSTMs. On the other hand, the NLSS-QoE proposed in [33] is a nonlinear QoE prediction model based on the conventional state space approach. It is shown in [33] that the NLSS-QoE model provides superior QoE prediction over the existing models on the LIVE Netflix [7] and the LFOVIA QoE [4] Databases.…”

“…Further, a continuous QoE prediction model is proposed based on Support Vector Regression (SVR). In [33], a QoE prediction model based on nonlinear state space (NLSS-QoE) has been proposed. [6] presents the Time-Varying QoE (TV-QoE) Indexer for predicting the continuous QoE using multi-stage and multi-learner HW approaches.…”