Mean Opinion Score

Description

The Mean Opinion Score (MOS) is a fundamental quantitative measure of the perceived quality of a transmitted speech, audio, or video signal as experienced by the end-user. It is defined in the ITU-T P-series recommendations and extensively adopted and elaborated within 3GPP specifications for service quality benchmarking. The core methodology involves a subjective listening or viewing test where a panel of human subjects rates the quality of processed audio/video samples under controlled conditions. Each subject provides an opinion score on an absolute category rating scale, typically: 5=Excellent, 4=Good, 3=Fair, 2=Poor, 1=Bad. The MOS is the arithmetic mean of all scores collected for a given test condition, resulting in a single number between 1.0 and 5.0.

The process for obtaining a MOS is highly standardized to ensure reproducibility and comparability. Test subjects are selected based on specific criteria (e.g., normal hearing), and they listen to or view sequences of processed media in a quiet room using standardized equipment. The test material includes clean source samples and samples that have been degraded by a codec, packet loss, delay, jitter, or other network impairments. The subjects are not experts but represent typical users. The rigorous control of environmental, equipment, and procedural variables is essential to isolate the impact of the system under test on perceived quality. The resulting MOS provides a direct, human-centric link between technical network parameters and user satisfaction.

Within the 3GPP architecture, MOS is not a directly measurable network Key Performance Indicator (KPI) but is the ultimate benchmark against which objective prediction models are calibrated. 3GPP specs define performance requirements for codecs, network delay, packet loss robustness, and other factors with the goal of achieving a target MOS under specific conditions. For example, a specification may require that a voice codec achieves a MOS of 4.0 or higher under a defined packet loss scenario. Network planning, codec selection, and QoS parameter tuning are all ultimately aimed at maximizing the MOS for the end-user. It bridges the gap between engineering metrics (e.g., latency, jitter) and the subjective human experience, making it indispensable for service quality management and standardization.

Purpose & Motivation

MOS was created to solve the fundamental problem of quantifying something inherently subjective: the quality of a communication experience. Before its standardization, comparing voice codecs or network performance was difficult and often based on inconsistent, proprietary listening tests. Engineers needed a reliable, repeatable method to evaluate how technical impairments (like bandwidth limitation, compression, or packet loss) actually impacted user perception. The MOS provided this common currency for quality.

Its adoption by 3GPP and other standards bodies was motivated by the need to define minimum performance requirements for digital cellular systems (GSM, 3G, 4G, 5G). As networks evolved from analog to digital, and introduced lossy compression codecs to save bandwidth, it became critical to ensure that these technological advances did not degrade voice quality below an acceptable threshold. MOS provided the definitive scale to set these thresholds (e.g., 'toll quality' is often associated with a MOS ~4.0). It allows for the objective comparison of completely different technological approaches (e.g., AMR-NB vs. EVS codec) based on their ultimate impact on the human listener, driving continuous improvement in speech and audio quality for mobile services.