Bidirectional LSTM for continuously predicting QoE in HTTP adaptive streaming

Tho Nguyen Duc, Chanh Minh Tran, Phan Xuan Tan, Eiji Kamioka

研究成果: Conference contribution

1 被引用数 (Scopus)

抄録

Continuous monitoring of quality of experience (QoE) has increasingly become an important mechanism in quantifying the user's satisfaction over video streaming services. Such mechanism requires approaches with the ability to model the complex temporal dependencies and time-varying characteristics of the QoE. Long-Short Term Memory (LSTM) related approaches carried out by recent research efforts have shown highly potential results since they can leverage past events occurring during a streaming session. However, the unidirectional LSTM structure only consider forward dependencies, leading to high possibility of missing out useful information. In this paper, a novel model for continuous QoE prediction is proposed to consider not only forward dependencies but also backward dependencies. The proposal utilizes a Bidirectional Long Short Term-Memory (BLSTM) model to process inputs obtained from perceptual video quality algorithms, rebuffering, and memory-related temporal data. Comparisons with other state-of-the-art models indicate that the proposed model achieves very promising performance in terms of accuracy.

本文言語English
ホスト出版物のタイトルACM International Conference Proceeding Series
出版社Association for Computing Machinery
ページ156-160
ページ数5
ISBN(印刷版)9781450361033
DOI
出版ステータスPublished - 2019
イベント2nd International Conference on Information Science and System, ICISS 2019 - Tokyo, Japan
継続期間: 2019 3 162019 3 19

出版物シリーズ

名前ACM International Conference Proceeding Series
Part F148384

Conference

Conference2nd International Conference on Information Science and System, ICISS 2019
国/地域Japan
CityTokyo
Period19/3/1619/3/19

ASJC Scopus subject areas

  • ソフトウェア
  • 人間とコンピュータの相互作用
  • コンピュータ ビジョンおよびパターン認識
  • コンピュータ ネットワークおよび通信

フィンガープリント

「Bidirectional LSTM for continuously predicting QoE in HTTP adaptive streaming」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル