Speech intelligibility is the result not of a single parameter but of an interaction of parameters. Loudspeaker coverage and aim determine the direct field, room volume and absorption determine the reverberant field, and in between, speech level and ambient noise determine how much of the original signal survives transmission when the human auditory system itself imposes limits at both ends of the level range. Improve one parameter while another dominates, and the STI barely moves.
This makes STI a difficult metric to optimize. Reading a single value tells you where you are, but it does not tell you which parameter is holding you back. Determining that usually involved setting up a calculation, running it, opening the stored results, changing a variable, and starting all over again.

EASE 5.84 eliminates most of that overhead. STI is now available as a metric in the Acousteer's real-time engine, mapped to audience areas directly in the main 3D view, and automatically recalculated whenever the model or a calculation parameter change. The calculation follows the same quality standard you are used to and it is made available at the speed of a design decision.

How the room reached the engine

Acousteer was introduced in EASE 5 Third Edition in November 2024 with only direct SPL mapping, based on a hybrid CPU/GPU implementation of the EASE simulation core. It covered the entire audio bandwidth, from 20 Hz to 20 kHz, with an internal spectral resolution of at least 1/24 of an octave, and every property of the model was hard-wired to the engine, so any modification triggered a recalculation.

Since then, key acoustic metrics have been added one by one, maintaining the same direction: moving away from the sound system in isolation and toward the sound system in the room. Arrival time and precedence effects were incorporated into the time domain. The total sound pressure level (total SPL), published in October 2025, introduced the statistically derived reverberant field. The D/R ratio and critical distance clarified the balance between direct field and reverberation, and the signal-to-noise ratio (S/N) added the ambient noise level.

STI is a milestone conclusion for a structural reason. It consumes all the previous elements. A real-time STI map requires that the direct field, statistical reverberant field, speech spectrum, and noise spectrum be present, consistent, and interact simultaneously within the same engine. Version 5.84 marks the point at which this becomes a reality.

What real-time STI contains

The implementation follows the indirect calculation method of the STI according to IEC 60268-16:2020. Direct field contributions are based on the CDPS model and calculated from complex GLL directivity data. The reverberant field is obtained statistically using the selected Eyring or Sabine model, or from a measured or reference reverberation time. 

Two level-dependent mechanisms complete the calculation, and each can be turned on or off independently.

The noise masking feature accounts for the effective signal-to-noise ratio caused by the ambient noise level. For high intelligibility, a signal-to-noise ratio of approximately 15 dB or higher is generally desirable; below this range, signal intelligibility degrades noticeably. Turning off noise masking reveals how much of the current result is solely due to the noise assumption.

The signal masking feature accounts for the loss in intelligibility that occurs when the level of the reproduced speech is too low or too high. As a practical guideline, the broadband voice level should not be significantly lower than 40 dB or higher than 90 dB when high intelligibility is required. Disabling signal masking removes the influence of the system’s operating level.

Both switches are analysis tools. Results intended to demonstrate compliance with IEC 60268-16:2020 must use the standardized voice input signal with both masking effects active. 

A tale of two switches

Taken together, the two masking switches turn an observation into an assessment.

If disabling noise masking changes the intelligibility map significantly, the project is noise-limited, and no adjustment will change that. If disabling signal masking makes a major impact on the intelligibility map, the system is operating outside the usable level range, and the correction likely depends on the gain structure, not the coverage. When neither switch produces a significant change, what remains is a coverage or room problem, and those are precisely the variables that Acousteer allows you to manipulate interactively.

The workflow that follows will be familiar to anyone who has manually optimized a system, with the difference that the feedback is immediate. A loudspeaker can be repositioned to check whether the improvement is applied to weak areas or simply reinforces an area that already meets the target. Gain, delay, and filters can be adjusted in a processing block, individual loudspeakers can be turned on, turned off, or isolated to attribute the result to specific parts of the system, material assignments can be reviewed to see whether the treatment would provide a significant improvement or just a decimal point. By raising the assumed noise spectrum to the worst-case scenario required by the specification, it is immediately shown which seats fall out of compliance first.

All of this happens directly in the 3D view and remains viable for systems with dozens or hundreds of speakers.

Interpreting the map along with the statistics

A map shows the spatial pattern of an outcome. It doesn't indicate what percentage of the audience meets the target or how uniform the actual coverage is. Therefore, Acousteer combines the 3D map with a real-time distribution graph that covers percentage of occurrence by class, along with the average, standard deviation, minimum, maximum, and number of data points. This data is recalculated in real time along with the map and gives you a robust reporting surface.

The research that supports the metric

While the groundbreaking research of Steeneken and Houtgast created the foundation for modern speech intelligibility assessment using the STI, its implementation in the shape of the IEC 60268-16 standard made it available to practitioners, equipment manufacturers, and authorities worldwide. 

In the context of acoustic simulation, work by Wolfgang Ahnert, Stefan Feistel and colleagues has focused on expanding STI from a reporting metric to a tangible design objective. 

Papers published by Ahnert and Feistel revolved around advancing calculations and software-based design principles, such as in [1] and [2]. Practical applications discussed, for example, with Anselm Goertz [3], Daniele Ponteggia [4], or Tim Kuschel [5] were concerned with beam-steering solutions improving speech intelligibility, optimizing listening conditions with respect to signal level, and managing acoustically adverse conditions such as in road tunnels. Additional background research with acoustic authorities like Michael Vorländer [6] was aimed at improving room-acoustic simulation and its input parameters in general. 

Version 5.84 applies this research to everyday interaction, making effects visible and shortening the feedback loop so that more time can be spent on technical analysis.

Configuration for a standard-compliant result

For results to demonstrate compliance, four conditions must be met:

1. The input signal is set to "Speech Critical" (IEC 60268-16:2020) in the Acoustic Parameters window.
2. Both noise masking and signal masking are active.
3. Ambient noise is defined as a spectrum representative of the operating conditions, measured on-site or extracted from a suitable NC/RC curve.
4. The reverberation base is adjusted to the characteristics of the room, and the statistical method is documented along with the result. 

Now available in EASE 5.84

Real-time STI mapping is included in EASE 5.84 for all active licenses. 

Current users can download the update from the AFMG download portal . 

New users can request a trial version of EASE 5 and see the feature first hand.

Further reading

1. Stefan Feistel; Wolfgang Ahnert; Alex Miron; Oliver Schmitz: ICA Madrid 2007: Improved Methods for Calculating Room Impulse Responses with EASE 4.2 AURA
2. Stefan Feistel; Wolfgang Ahnert: POMA 2025: Complex acoustic simulations in real time
3. Stefan Feistel; Anselm Goertz: AES New York 2013: Digitally Steered Columns: Comparison of Different Products by Measurement and Simulation
4. Stefan Feistel; Daniele Ponteggia: AES Le Mans 2024: Optimizing Speech Intelligibility in Practical Applications using State-of-the-Art Acoustic Simulation Tools
5. Stefan Feistel; Tim Kuschel: AES New York 2024: Acoustic modeling and designing of emergency sound systems in road tunnels
6. Michael Vorländer; Stefan Feistel: POMA 2024: Show your scattering coefficients