How to Measure Speech Intelligibility

6 steps15-20 min readUpdated 2026-03-20

Quick Answer

Measuring speech intelligibility means quantifying how clearly speech can be understood through an electroacoustic system or in a room using the Speech Transmission Index (STI). Defined by IEC 60268-16, STI ranges from 0 (unintelligible) to 1 (perfect), derived from the modulation transfer function across 7 octave bands.

Try It Now

Open STI Measurement Tool — free, no download required.

Open Tool

Equipment Needed

  • Omnidirectional measurement microphone
  • STIPA test signal source (built into SonaVyx)
  • Loudspeaker system under test
  • SonaVyx STI measurement tool
  • SPL meter for background noise measurement

Step-by-Step Guide

1

Select Measurement Method

Choose between STIPA (simplified, 15-second test signal) and full STI (derived from impulse response). STIPA uses a pre-recorded test signal with specific modulation frequencies per IEC 60268-16 Table F.1 and is the most common field method. Full STI derived from the room impulse response provides higher accuracy and allows prediction of intelligibility changes. SonaVyx supports both methods. For emergency voice alarm systems, STIPA is typically the required method per EN 54-16.

2

Configure the Sound Source

Play the STIPA test signal through the loudspeaker system at normal speech operating level, typically 60 to 65 dBA at 1 meter from the talker position or the nearest speaker. For distributed ceiling speaker systems, measure the combined output of all relevant speakers. The source level should represent realistic speech reinforcement conditions. Do not use excessive levels that would not occur during actual use, as this inflates the measured STI by improving the signal-to-noise ratio.

3

Position Measurement Microphone

Place the omnidirectional measurement microphone at seated ear height (1.2 meters) or standing ear height (1.5 meters) at the listener position being assessed. For certification measurements, IEC 60268-16 requires measurements at multiple positions distributed across the coverage area. Common practice is to measure on a grid with points every 3 to 5 meters, or at designated worst-case positions (rear corners, under balconies, near noise sources).

4

Capture and Analyze

Start the measurement in SonaVyx's STI tool. For STIPA, the tool analyzes the received signal for at least 15 seconds (Quick mode) or three averaged 15-second captures (Pro mode). The analysis extracts the modulation transfer function (MTF) at 14 modulation frequencies across 7 octave bands (125 Hz to 8 kHz). Low MTF values indicate where the system fails to preserve speech modulations, pointing to the specific frequency range and cause of intelligibility problems.

5

Interpret the Results

STI above 0.75 is rated Excellent (A grade). STI from 0.60 to 0.75 is Good (B). STI from 0.45 to 0.60 is Fair (C). Below 0.45 is Poor to Bad. For most speech-reinforcement applications, the minimum acceptable STI is 0.50. Emergency voice alarm systems typically require STI of 0.50 or higher per EN 54-16. Worship spaces should target at least 0.55. If STI falls below requirements, the per-band MTF data helps diagnose whether reverberation, noise, or system response is the primary degradation mechanism.

6

Identify and Address Problems

Low MTF in low bands (125 to 500 Hz) usually indicates excessive reverberation, which reduces STI by smearing speech modulations over time. Low MTF in high bands (2 to 8 kHz) suggests inadequate high-frequency response or high background noise masking consonants. SonaVyx displays per-band MTF charts and identifies the primary degradation factor. Solutions include adding absorption to reduce RT60, improving the signal-to-noise ratio by reducing background noise or increasing system level, and redesigning speaker coverage to reduce late-arriving reflections.

Understanding the Modulation Transfer Function

Speech intelligibility depends on the listener's ability to follow the amplitude modulations that encode consonants and syllable structure. The MTF measures how faithfully these modulations are transmitted from source to receiver at each frequency. An MTF of 1.0 means modulations are perfectly preserved; 0.0 means they are completely lost. Reverberation, background noise, distortion, and echoes all reduce the MTF through different mechanisms.

STI Computation from MTF

The STI calculation first converts each of the 98 MTF values (14 modulation frequencies times 7 octave bands) to an apparent signal-to-noise ratio using a logarithmic transformation. These values are clipped to a range of -15 to +15 dB. The modulation frequencies within each octave band are averaged, then the 7 band values are combined using male speech weighting factors from IEC 60268-16 Table 4. Redundancy correction factors account for correlations between adjacent octave bands.

STIPA vs Full STI

STIPA (Speech Transmission Index for Public Address) is a simplified method that uses only 2 modulation frequencies per octave band instead of 14, reducing the test signal duration from several minutes to 15 seconds. IEC 60268-16 Ed.5 demonstrates that STIPA correlates within plus or minus 0.03 of full STI for typical electroacoustic systems. STIPA cannot detect certain non-linear distortions that affect only specific modulation frequencies, but this limitation rarely matters in practice.

Factors Affecting STI

Three primary factors degrade STI: reverberation (RT60 above 1.0 seconds significantly impacts intelligibility), background noise (SNR below 15 dB progressively reduces STI), and system frequency response (inadequate high-frequency coverage above 2 kHz removes consonant clarity). The relationship between RT60 and STI is approximately STI = 0.9 minus 0.1 times RT60 for rooms with adequate SNR, providing a useful estimation when measurements are not possible.

Common Mistakes to Avoid

Playing the STIPA signal through multiple speakers simultaneously when each speaker zone should be tested independently

Measuring STI with the room empty, which underestimates the occupied RT60 and overestimates real-world intelligibility

Setting the test signal level unrealistically high, producing artificially good STI from improved SNR

Confusing CIS (Common Intelligibility Scale) with STI, which use different numerical scales

Ignoring background noise measurement, which is needed to separate reverberation effects from noise effects on MTF

Applicable Standards

StandardClauseRelevance
IEC 60268-16:2020Clause 4Complete STI computation methodology from modulation transfer function
IEC 60268-16:2020Annex FSTIPA test signal specification and measurement procedure
EN 54-16:2008Clause 5STI requirements for voice alarm systems in buildings

Try It Now

Open STI Measurement Tool — free in your browser.

Open Tool

Related Guides

Design-Time Companion

Need to design room acoustics before measuring? AcousPlan handles room acoustic design, while SonaVyx handles deploy-time measurement and tuning.

Frequently Asked Questions