Reliable Adaptive Video Streaming Driven by Perceptual Semantics for Situational Awareness

Area of Science:

• Computer Science
• Information Theory
• Network Engineering

Background:

• Existing video adaptation solutions are insufficient for live streaming in challenging communication environments.
• Recreational video streaming quality metrics do not apply to critical applications requiring enhanced situational awareness.
• Need for a robust framework to manage video quality under network constraints.

Purpose of the Study:

• To present a novel cross-layer optimized video adaptation technique driven by perceptual semantics.
• To enhance situational awareness through improved live video streaming under adverse network conditions.
• To introduce a new Quality of Experience (QoE) framework for controlled video adaptation with perceptual feedback.

Main Methods:

• Developed a novel Quality of Experience (QoE) framework enabling controlled adaptation and perceptual semantic feedback.
• Employed temporal/spatial abstraction for video applications beyond recreational use.
• Utilized cross-layer optimization integrating network congestion (time) and erasure (space) feedback for bandwidth distribution.
• Implemented Systematic Random Linear Network Coding (SRNC) for enhanced reliability while preserving perceptual semantics.

Main Results:

• Objective metrics demonstrated homogeneous high performance for perceptual features using the proposed scheme.
• The cross-layer optimization effectively managed scarce bandwidth by considering network feedback.
• SRNC successfully improved video stream reliability without compromising perceptual quality.

Conclusions:

• The proposed video adaptation scheme significantly enhances live streaming quality for situational awareness.
• The novel QoE framework provides effective control and feedback for perceptually optimized video delivery.
• The approach aligns with content-aware networking principles and information-centric-networking architectures.