Non-Intrusive MOS Estimation via Packet Loss Polynomial Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for evaluating voice quality in packet loss environments are intrusive, making them computationally expensive and unsuitable for non-intrusive measurement devices, which is a limitation in predicting customer opinion based on actual network performance.
Innovation Solution
A non-intrusive method and system for estimating mean opinion score (MOS) using packet loss patterns, involving a polynomial function that maps packet loss probabilities to MOS scores, with coefficients determined through least squares analysis of reference scores from intrusive algorithms, allowing for scalable and accurate voice quality analysis without accessing packet contents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If intrusive algorithms (PESQ) are used to evaluate voice quality, then measurement precision is improved, but device complexity and computational cost increase
Solution Approach 1:
The patent creates a simplified copy of the intrusive PESQ algorithm by developing a non-intrusive polynomial model that replicates MOS score predictions. Instead of implementing the full intrusive algorithm, a polynomial function with pre-determined coefficients is used to copy the essential prediction capability while avoiding packet content extraction and reassembly requirements.
Solution Approach 2:
The patent replaces the computationally expensive intrusive algorithm with a lightweight polynomial calculation that requires minimal processing resources. The complex PESQ algorithm is substituted with a simple polynomial function evaluation that can be performed rapidly using only packet loss probability as input, significantly reducing computational cost.
2Measurement precision
If intrusive algorithms are used to access packet contents for quality assessment, then measurement precision is improved, but productivity and scalability deteriorate
Solution Approach 1:
The patent extracts only the essential input parameter (packet loss probability) from the packet data stream while discarding the need to extract and process packet contents. By taking out only the necessary information (packet loss metrics) and eliminating the requirement to access packet payloads, the system achieves both accuracy and high processing speed.
Solution Approach 2:
The patent segments the quality assessment process into two independent parts: (1) packet loss measurement from packet headers, and (2) MOS score calculation using the polynomial model. This segmentation allows packet loss tracking to be performed independently without requiring packet content extraction, enabling parallel processing and improved productivity.
3Measurement precision
If packet contents are extracted and reassembled for algorithm processing, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary action by pre-determining the polynomial coefficients through least squares analysis of reference MOS scores during an offline training phase. This preliminary preparation allows the online evaluation to直接使用 the pre-computed model without requiring real-time packet reassembly or complex processing, significantly reducing evaluation delay.
4Productivity
If non-intrusive measurement is implemented without packet content access, then productivity and scalability are improved, but measurement precision deteriorates
Solution Approach 1:
The patent changes the measurement parameters by using packet loss probability as the sole input parameter instead of requiring full packet content analysis. The polynomial model is specifically designed to map packet loss probability directly to MOS scores, transforming the measurement approach to achieve both non-intrusive operation and acceptable accuracy for network performance monitoring.
Data Source
AI summary
Methods, systems, and computer readable media for non-intrusive mean opinion score (MOS) estimation based on packet loss pattern are disclosed. According to one aspect, a method for non-intrusive mean opinion score estimation based on packet loss pattern includes receiving a packet data stream, measuring the packet loss for the received data stream, calculating a probability of packet loss based on the measured packet loss, and calculating an estimated mean opinion score based on the calculated probability of packet loss. In one embodiment, the estimated mean opinion score is calculated using a mathematical function that maps calculated probability of packet loss to mean opinion score. In one embodiment, the mathematical function is a polynomial having coefficients that are selected so that the polynomial closely models reference opinion scores for a range of packet loss probabilities. In one embodiment, the coefficients are determined using a least squares analysis of a dataset that includes reference opinion scores for each of a range of packet loss probabilities. In one embodiment, the reference opinion scores for each of a range of packet loss probabilities are calculated using an intrusive algorithm to analyze lossy data streams that exhibit particular packet loss probabilities and generate mean opinion scores for each of the respective packet loss probabilities.