Listening LABoratory

Description

The Listening LABoratory (LAB) is a formalized test methodology and environment specified in 3GPP TS 26.935. Its primary function is to assess the subjective quality of speech and audio codecs used in mobile communications. The LAB is not a physical piece of network equipment but a rigorous procedural framework. It defines the entire chain for quality evaluation, including the selection of expert listeners, the creation of test sequences using standardized source material, the specific conditions for playback and listening, and the statistical methods for analyzing listener scores. This process generates a Mean Opinion Score (MOS), which is a critical metric for codec performance.

The architecture of a LAB is centered around a controlled acoustic environment. Listeners are placed in soundproof booths with calibrated audio equipment to eliminate external noise and ensure consistent playback levels. The test sequences are carefully crafted to represent a wide range of acoustic conditions and include processed speech (through the codec under test) alongside hidden reference signals. The core methodology involves presenting these sequences to a panel of listeners who rate the perceived quality on a standardized scale. The resulting scores are then processed using specific statistical techniques outlined in the specification to produce reliable and comparable MOS values.

The role of the LAB in the 3GPP ecosystem is foundational for codec standardization. Before a new speech or audio codec (like AMR, EVS, or AMR-WB) is approved for inclusion in the specifications, it must undergo rigorous LAB testing to prove it meets minimum quality thresholds and offers improvements over existing codecs. This ensures that technological advancements in compression and bandwidth efficiency do not come at the cost of unacceptable audio quality for the end user. The LAB provides the objective, reproducible evidence needed to make these standardization decisions.

Purpose & Motivation

The LAB was created to solve the fundamental problem of how to objectively and reliably measure the subjective experience of speech quality in digital mobile networks. Early digital codecs introduced artifacts like quantization noise and bandwidth limitation, which were difficult to quantify with simple electrical measurements. The purpose of the LAB is to provide a standardized, scientific method for translating human perception into a quantifiable metric (MOS), enabling fair comparison between different codec algorithms and implementations.

Historically, without a standardized listening test method, codec developers and network operators could not reliably compare results from different testing facilities. This made the codec selection and standardization process subjective and potentially biased. The LAB specification addresses this by defining every controllable variable—from the physical listening environment and listener screening to the test material and statistical analysis—ensuring that results are reproducible and comparable across different laboratories worldwide. It was motivated by the need for a common 'language' of quality assessment to drive the evolution of speech codecs from GSM's Full Rate to today's Enhanced Voice Services (EVS) and beyond, always with a verified user-centric quality benchmark.

Key Features

• Standardized test methodology for subjective speech and audio quality assessment
• Definition of controlled acoustic environment and listener screening criteria
• Procedure for generating and presenting test sequences including processed and reference signals
• Statistical processing of listener scores to calculate a Mean Opinion Score (MOS)
• Ensures reproducibility and comparability of test results across different laboratories
• Provides the foundational quality evidence required for 3GPP codec standardization

Evolution Across Releases

Defining Specifications