Initial Bitrate Prediction for Stable Real-Time Communication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing real-time communication systems face challenges in determining an optimal initial bitrate due to varying user and network conditions, leading to suboptimal quality and degradation issues such as delays and freezes.
Innovation Solution
An initial bitrate determiner uses an initial bitrate predictor model trained on historical communication data to generate a probability distribution for determining an optimal bitrate based on geographical location, network type, and previous communication statistics, minimizing the risk of bandwidth overshooting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed predetermined bitrate is used for real-time communication, then the system is simple to implement, but the communication quality degrades under varying network conditions
Solution Approach 1:
The patent implements dynamic bitrate determination by training a machine learning model on historical communication data to predict optimal initial bitrates based on various features. The model adapts to different network conditions, user devices, and geographical locations, replacing the static fixed bitrate approach with a dynamic prediction system that adjusts bitrate based on learned patterns from training data.
Solution Approach 2:
The patent applies preliminary action by training the bitrate predictor model in advance using extensive historical communication records. The model learns optimal bitrate patterns beforehand, enabling it to make accurate predictions when establishing new communications without requiring real-time network measurements or complex handshaking protocols.
2Reliability
If a high bitrate is selected to improve communication quality, then the quality increases, but delays and freezes occur when the bitrate exceeds available bandwidth
Solution Approach 1:
The patent incorporates feedback by training the machine learning model on historical communication data that includes actual communication outcomes. The model learns from past experiences what bitrates led to successful communications without delays versus those that caused buffering and freezes, enabling it to predict optimal bitrates that balance quality and timeliness.
Solution Approach 2:
The patent changes the bitrate parameter dynamically based on predicted network conditions and historical patterns. Instead of using a fixed or uniformly high bitrate, the system adjusts the initial bitrate parameter according to the trained model's predictions, which consider multiple features including network type, device characteristics, and geographical location to optimize both quality and avoid delays.
3Ease of operation
If a fixed bitrate is used regardless of network conditions, then the system is easy to implement, but bandwidth utilization is inefficient
Solution Approach 1:
The patent implements self-service by enabling the system to automatically determine optimal bitrates without manual configuration. The trained machine learning model autonomously predicts the best initial bitrate based on input features, eliminating the need for network administrators to manually tune bitrate settings for different conditions while maximizing bandwidth utilization efficiency.
Solution Approach 2:
The system dynamically changes the bitrate parameter based on predicted conditions and historical patterns, optimizing bandwidth utilization automatically. The model learns from training data what bitrates achieve the best throughput and quality for different network scenarios, enabling efficient bandwidth usage without requiring complex manual configuration or real-time network measurements.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A method (600) for determining an initial bitrate (180) for a communication includes receiving a communication request (21) to establish a digital communication between a first user device (10a) and a second user device (10b) associated with a plurality of features (310) including a geographical identifier identifying a geographical location associated with the first user device, a first network type connection associated with the first user device, a second network type connection associated with the second user device, and an average bitrate for a previous digital communication of the first user device. The method includes determining, using an initial bitrate predictor model (160) configured to receive the plurality of features as feature inputs, an initial bitrate for the digital communication between the first user device and the second user device, and establishing the digital communication between the first user device and the second user device at the determined initial bitrate.