A QoE aware LSTM based bit-rate prediction model for DASH video

Dynamic Adaptive Streaming over HTTP (DASH) is becoming the de facto technology for live and on-demand video streaming services. This is primarily due to the flexibility it offers toward delivering satisfactory Quality of Experience (QoE) to a variety of end users over temporally varying unpredictable wireless channels. Although, the DASH framework provides the flexibility for selecting an appropriate bit-rate for each segment, however, it is still a challenging task to deliver a satisfactory QoE to the end user over entire video session. In this work, we present an efficient bit-rate prediction algorithm for the DASH framework that attempts to deliver high quality uninterrupted smooth video viewing experience to the end user. We have formulated the prediction model using Long Short-Term Memory (LSTM) due to its ability to provide good prediction output results in sequence learning. Experimental results reveal that the proposed prediction model is able to predict the bit-rate of the segments effectively and thus, deliver satisfactory QoE even in the face of highly varying channel conditions.