Bit Rate Prediction Neural Network for Adaptive Video Playback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for addressing video picture freezing during multimedia data playback, such as reducing bit rate based on network conditions, result in poor smoothness and increased lag, leading to suboptimal user experience.
Innovation Solution
A multimedia data processing method that employs a bit rate prediction neural network to predict accurate bit rates, using classification and regression learning to generate a second bit rate prediction model with low running overhead, enabling adaptive bitrate adjustments in real-time to prevent freezing and improve playback smoothness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If bit rate is reduced based on current network conditions, then video freezing is alleviated, but playback smoothness deteriorates due to lag
Solution Approach 1:
The patent applies preliminary action by predicting future network conditions and proactively adjusting bit rate before actual network changes occur. The bit rate prediction neural network analyzes historical network data and current status to forecast upcoming network conditions, allowing the system to pre-adjust bit rate settings. This prevents video freezing in advance while avoiding reactive adjustments that would cause playback lag, as the system acts before the network condition actually changes rather than responding after the fact.
2Measurement precision
If complex neural network is used for bit rate prediction, then prediction accuracy is improved, but terminal running overhead increases
Solution Approach 1:
The patent applies segmentation by dividing the bit rate prediction system into two distinct parts: a bit rate prediction neural network that performs complex analysis and generates prediction results, and a separate bit rate prediction model that is deployed on the terminal. The neural network processes comprehensive network data to create accurate predictions, while the simplified model on the terminal executes these predictions with minimal computational overhead. This segmentation allows the terminal to benefit from high prediction accuracy without bearing the full computational burden of a complex neural network.
Data Source
AI summary
A multimedia data processing method includes: acquiring a first predicted bit rate predicted by a bit rate prediction neural network based on first network data; performing classification and regression learning on the first network data and the first predicted bit rate to obtain a first bit rate prediction model; acquiring second network data and a second predicted bit rate predicted by the first bit rate prediction model, and the second network data being collected during playback of multimedia data using the second predicted bit rate; and performing classification and regression learning on the first bit rate prediction model based on a difference between the first network data and the second network data and a difference between the first predicted bit rate and the second predicted bit rate, to obtain a second bit rate prediction model configured to perform bit rate prediction in a terminal that plays the multimedia data.