Quantifying Speech Intelligibility

Online Editor

Brian Johansen explains how improving human-vehicle communication can create a safer and less stressful environment for travellers

People spend a lot of hours in transit, sometimes travelling alone or in close proximity to others. These experiences can potentially have a considerable impact on our sense of well-being, so there is a lot of focus on keeping the cabin calm and quiet.

A more recent in-cabin issue stems from the increasing prevalence of infotainment systems that provide hands-free alternatives to looking away from the road and taking hands of the wheel. It is important to ensure that when there is communication, from passenger to passenger or operator to infotainment system, that communication is clear. This means that much attention must be paid  to supressing the unwanted noise from the engine, traffic, wind and tires. Along with supressing this unwanted noise, care must be made to ensure that acoustic treatments don’t dampen or muddle conversation and the ability to interact with phones or infotainment systems.

GRAS’ 45BB-4 KEMAR Head & Torso Simulator (HATS) for sound quality recording is designed with realistic head and body geometry to provide realistic and quantifiable data from the driver and passenger seats to ensure clear communication in a nice, calm environment.

To quantify the data and create a situation where the data can be objectively evaluated, the Speech Transmission Index (STI) is needed. The STI is a measurement and calculation method that includes a number of parameters. Using the STI along with valid data, one can calculate the speech intelligibility that will be achieved under the given circumstances. The STI finding is presented on a scale and is used in various standardised norms as an expression of how understandable a speech is received. A higher value means a better understanding – less effort for the recipient – which can contribute to a better user experience. However, a key factor in usefully applying this index is accurate, valid data that matches realistic communication in a cabin. The best way to acquire such data is through the use of a HATS such as GRAS KEMAR, which is specifically designed to acquire the data just as an actual human occupying that space would.

How Gras KEMAR Works

KEMAR meets the IEC 60318-7 human head and ear simulator standard that describes a realistic head and body geometry. With a HATS like KEMAR occupying the measurement positions, a realistic measurement environment is created because it fills the space in the cabin that otherwise consists of driver and passenger. This creates realistic sound reflections and dampening just as and when a person is sitting in a car. Using two or more KEMAR in the same cabin, you can simulate how a talker and a receiver interact.

KEMAR’s binaural capability – microphone in both the left and right ear – provides a benefit beyond the basic acquired STI data. For example, the ear closest to the window will typically perceive more wind noise and thus reduce the general speech intelligibility. Being able to separate the right and left channels in your research can be useful for pin-pointing where improvements can be made. If there is a 3D geometry of the HATS (as there is with KEMAR), that performance can be simulated in FEM/BEM models, already in the design phase.

A Full Test System

Further from the STI, specifically concerned with human communication with infotainment systems, are the Perceptual Objective Listening Quality Assessment (POLQA) and Perceptual Evaluation of Speech Quality (PESQ). These tests use the cabin environment and a speaker and receiver to compare an audio output (at the phone line’s ‘listener’ end) to the original voice file (played on the ‘talker’ side) to create a fully unbiased and objective indicator of the actual audio being heard. This is more accurate than the other methods of audio quality measurement which often rely on audio quality predictions based on network performance.

Combining an analyser and software that support POLQA and PESQ, such as Audio Precision APx500 with KEMAR – which can also be used for measuring the transmission performances of car hands-free terminals – can form a full test system that provides the required data to improve speech intelligibility and overall human-vehicle communication earlier in the design process. These tools together can create a safer, less stressful and more pleasing environment for vehicle occupants.

Brian Johansen is an Acoustic Engineer at GRAS Sound & Vibration