Encrypted OTT Video Quality Analysis via RNN Packet Probes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for monitoring the quality of over-the-top (OTT) video services in mobile networks are limited by the use of different video encoding techniques and the inability to sample encrypted content, leading to inefficient fault identification and troubleshooting.
Innovation Solution
A real-time video quality analytics system based on a recurrent neural network (RNN) that adapts to various video encodings and provides fine-granularity metrics for encrypted OTT data, using packet probe records and network node events to identify and troubleshoot issues automatically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traffic probes are used to collect metrics for monitoring OTT video quality, then video quality metrics can be obtained, but the system cannot handle encrypted content and requires separate tuning for each encoding technique
Solution Approach 1:
The patent introduces an intermediary processing layer that sits between the encrypted video traffic and the quality analysis system. This intermediary uses machine learning models to infer quality metrics from encrypted packet characteristics without requiring direct access to the video content or knowledge of specific encoding techniques, thus maintaining both measurement precision and adaptability
Solution Approach 2:
The system changes from analyzing video content directly to analyzing packet-level parameters such as timing, size, and flow patterns. By transforming the measurement approach from content-based to metadata-based parameters, the system achieves universal compatibility across different encoding techniques while maintaining quality assessment capability
2Measurement precision
If terminal-based monitoring is deployed to monitor OTT flows, then quality metrics can be captured at the end device, but it requires subscriber agreement and consumes terminal processing resources
Solution Approach 1:
The patent enables the network infrastructure to perform quality monitoring autonomously without requiring end-user participation. Network probes and analytics systems automatically capture and analyze packet characteristics, eliminating the need for subscriber agreement or terminal-based software deployment while maintaining accurate quality measurement
3Measurement precision
If machine-learning-based techniques are used to infer network conditions, then quality metrics can be estimated, but the system requires access to network nodes and multiple data sources with significant parameter variation
Solution Approach 1:
The patent develops a universal machine learning framework that processes multiple data sources through a single standardized interface. The system accommodates variations in encoding techniques and network conditions by using a unified model architecture that automatically adapts to different input types, reducing system complexity while maintaining estimation accuracy
4Device complexity
If existing machine-learning techniques provide single metric per session, then analysis is simplified, but insufficient granularity is provided for troubleshooting specific issues
Solution Approach 1:
The patent segments the quality analysis into multiple granular components including per-segment video quality, network condition metrics, and temporal quality variations. This segmentation enables detailed troubleshooting of specific issues while maintaining organized, manageable data structures that do not excessively increase system complexity
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
Techniques for real-time encrypted over-the-top (OTT) data quality analysis are presented. For instance, the present disclosure includes an example method performed by a device (152) in a recurrent neural network (RNN) (150) implemented in a data communication system (10). The method includes periodically obtaining (302) information corresponding to one or more OTT data sessions (16) in the data communication system, the information being obtained from one or more of a packet probe (120), a test terminal (102T), and a network node (106). The method also includes generating (304) at least one quality metric associated with the one or more OTT data sessions (16) based on the periodically obtained information. Related devices, processor and memory arrangements, methods, and computer programs are also presented.