Degradation Mean Opinion Score

Description

Degradation Mean Opinion Score (DMOS) is a key performance indicator (KPI) used in 3GPP specifications to quantify the perceptual quality degradation of video services as experienced by end-users. Unlike traditional Mean Opinion Score (MOS), which provides an absolute quality rating, DMOS specifically measures the perceived difference, or degradation, between a reference (original) video sequence and a processed or transmitted version that has undergone impairments like compression artifacts, packet loss, jitter, or rebuffering.

The calculation of DMOS typically involves subjective testing methodologies standardized by bodies like ITU-T, where human viewers rate the impaired video relative to the reference on a scale. In network operations, objective models (algorithms) predict DMOS by analyzing video characteristics and network impairment parameters without human intervention. These models, such as those defined in ITU-T P.1203 or other VQM (Video Quality Metric) approaches, ingest metrics like bitrate, frame rate, packet loss rate, and stalling events to compute a predicted DMOS value.

Within the 3GPP architecture, DMOS is monitored and reported as part of the Quality of Experience (QoE) measurement collection framework defined for multimedia services. The UE or a network probe can measure media delivery performance and, using a standardized model, calculate DMOS. This data is then reported to the network, often to a Traffic Detection Function (TDF), Policy and Charging Rules Function (PCRF), or directly to the operator's management system via the Management Data Analytics (MDA) function.

The role of DMOS is integral to closed-loop network optimization. By tracking DMOS values in real-time or via aggregated reports, network operators can identify service degradation hotspots, correlate them with radio conditions or core network load, and trigger policy actions. For example, if DMOS for a video streaming service falls below a threshold in a certain cell, the network might prioritize that user's traffic or adjust radio resource allocation to improve the QoE, ensuring subscriber satisfaction and reducing churn.

Purpose & Motivation

DMOS was introduced to address the limitations of simple network-centric metrics (like throughput, latency, packet loss) in capturing the actual user perception of video quality. As mobile networks evolved to deliver high-bandwidth video services (streaming, conferencing), operators needed a way to measure success not just in technical delivery but in perceived quality. Traditional MOS provided an absolute score but didn't efficiently highlight the impact of network-induced degradations.

The creation of DMOS was motivated by the need for a differential metric that could sensitively reflect quality changes due to network conditions. It solves the problem of quantifying the 'annoyance factor' or quality drop a user experiences during events like video stalling, resolution switches, or compression artifacts. This allows for more targeted optimization; an operator can set thresholds on acceptable degradation levels and automate network responses. Historically, its adoption in 3GPP standards (starting in Release 8 for MBMS and later for general QoE monitoring) enabled a shift from best-effort service delivery to QoE-assured service delivery, which is fundamental for premium video services and competitive differentiation.

Key Features

• Measures perceptual video quality degradation relative to an unimpaired reference signal.
• Supports both subjective (human-rated) and objective (algorithm-predicted) measurement methodologies.
• Integrated into 3GPP's QoE measurement collection and reporting framework for multimedia services.
• Enables network-triggered policy actions based on real-time or historical QoE degradation thresholds.
• Used for benchmarking and optimizing video delivery performance across different radio access technologies (e.g., LTE, 5G NR).
• Facilitates correlation between application-layer quality (DMOS) and underlying network KPIs for root-cause analysis.

Evolution Across Releases

Defining Specifications