Skip to content

LESSON 99

Speech-in-Noise Tests: QuickSIN, HINT, Matrix

The most common complaint of people with hearing loss is understanding speech in noise, and a word score in quiet does not fully show this. In this lesson we look at signal-to-noise ratio and SNR loss, how QuickSIN, HINT and matrix tests are built, how to interpret their results and how to use them in hearing aid counseling.

Watch on YouTube

Follow along

Hello, and welcome to the AIHEAR Audiology Academy.
In this lesson, we learn about speech-in-noise tests and the idea of SNR loss that they measure.
We will see how QuickSIN, HINT and matrix tests are given and how their results are used in counseling.
The most common complaint of people with hearing loss is not being able to understand speech in crowded, noisy places.
A word recognition score measured in a quiet booth does not reliably predict performance in noise.
Two people with almost identical audiograms can understand speech in noise at very different levels.
So speech-in-noise tests are done in addition to pure tone and quiet speech tests.
The key measure in these tests is the difference between the speech level and the noise level.
This difference is called the signal-to-noise ratio, or SNR, and it is expressed in decibels.
Calculating SNR is simple: subtract the noise level from the speech level.
If speech is sixty five and noise is sixty decibels, the SNR is plus five decibels; speech is louder than the noise.
If the two are equal, the SNR is zero.
If noise is louder than speech, SNR is negative; young normal-hearing adults can still partly understand sentences in this condition.
Speech-in-noise tests measure how much SNR a person needs to understand half of the speech.
In the graph, the green curve shows how the percentage of words correct changes with SNR in normal-hearing people.
The red curve belongs to a sample patient who struggles in noise; a higher SNR is needed to reach the same percentage.
The difference between the fifty percent points of the two curves is called SNR loss; in this example, eight decibels.
So this patient needs speech to be eight decibels louder than the noise to understand as much as a normal-hearing listener.
The key point: raising both speech and noise together does not change the SNR, so it does not make up for SNR loss.
Reducing SNR loss needs solutions that separate speech from noise, such as directional or remote microphones.
In the fixed SNR method, sentences are presented at preset SNR values.
The result is the percentage of words repeated correctly, or a calculation based on these counts.
In the QuickSIN test we will see shortly, SNR drops in preset steps with each sentence; this is also a fixed scheme.
In the adaptive method, the SNR of the next sentence is set according to the patient's previous response.
If the patient answers correctly, the test gets harder; if wrong, easier; so it moves back and forth around the fifty percent point.
The result is given in decibels as the SNR at which fifty percent of the speech is understood, the SRT50.
The Quick Speech-in-Noise test, or QuickSIN, was published by Killion and colleagues in 2004.
The sentences are read by a female talker and presented in a babble noise of four people talking at once.
A list has six sentences, and five key words are scored in each sentence.
The first sentence is at plus twenty five decibels SNR; the SNR drops five decibels with each sentence, and the last is at zero SNR.
The key words the patient repeats correctly are counted; the maximum total is thirty.
SNR loss is found by subtracting the total number correct from twenty five point five.
If the average threshold is up to forty five decibels HL, the test is at seventy decibels HL; if worse, at a loud but comfortable level.
According to the test manual, an SNR loss of zero to three decibels is considered normal or near normal.
Three to seven decibels is a mild SNR loss.
Seven to fifteen decibels is a moderate SNR loss, and above fifteen is a severe SNR loss.
In this last group, understanding speech in noise is very hard, even with hearing aids.
An example: the patient repeated twenty of thirty key words correctly; twenty five point five minus twenty is five point five decibels.
This patient has a mild SNR loss.
The Hearing in Noise Test, or HINT, was published in 1994 by Nilsson, Soli and Sullivan.
The test uses short, everyday sentences; a sentence counts as correct only if all its words are repeated correctly.
Speech-shaped noise is presented at a fixed level, and the sentence level goes up or down depending on the patient's response.
Steps of four decibels are used for the first sentences, then steps of two decibels.
The result is given in decibels as the SNR at which half of the sentences are understood correctly.
In the sound field, noise can come from the front or the sides, to assess the benefit of hearing with two ears.
There is also a Turkish HINT, with a version for children.
The first matrix test was developed for Swedish; today comparable versions exist in German, Turkish and many other languages.
Sentences follow a fixed five-word pattern, and the words are picked at random from a ten-choice table, so they cannot be predicted.
The graph shows the track of one measurement: SNR falls as the patient answers correctly and rises after errors.
The next SNR step is set by the number of words repeated correctly in each sentence.
After twenty sentences, the track settles at the fifty percent point; this point is reported as the SRT50.
Since the score improves as the patient learns the word set, a practice list is given before the measurement.
SNR loss is used in hearing aid counseling to set realistic expectations with the patient.
If SNR loss is near normal, the person understands speech in noise close to normal-hearing listeners.
With a mild SNR loss, near-normal performance in noise can be expected with a hearing aid with directional microphones.
With a moderate SNR loss, directional microphones help clearly but may not be enough on their own.
A severe SNR loss needs the largest SNR improvement; remote microphone systems placed near the talker are considered.
These suggestions are a general framework; the decision is made with the patient's needs, audiogram and preferences.
In this example, the two patients' audiograms are almost identical: a mild to moderate loss sloping toward the high frequencies.
Their word recognition scores in quiet are also close to each other.
Patient A has an SNR loss of two decibels and performs near normal in noise.
Patient B has an SNR loss of twelve decibels and has moderate difficulty in noise.
Even with the same hearing aid, these two patients will have very different experiences in noise.
A speech-in-noise test makes this difference measurable and makes counseling personal.
Short tests have limited reliability; take this into account.
Its developers report that a single QuickSIN list has about plus or minus two point seven decibels of uncertainty at ninety five percent.
So averaging at least two lists before making a decision strengthens the result.
When comparing unaided and aided performance, use different but equivalent lists in each condition.
Sentence tests depend on language; the patient's language and vocabulary affect the result.
Holding sentences in mind takes working memory and attention, and in older patients this also shows in the result.
The result depends on whether it was measured with earphones or in the sound field, at what level and with which noise.
So each test is compared only with norms in its own language and for its own version.
The report should clearly state the test name, list, noise type, level and listening condition.
A word score in quiet does not reliably show how well a person understands speech in noise.
SNR loss shows how much more SNR a person needs than normal-hearing listeners to understand speech.
In QuickSIN, SNR loss is calculated by subtracting the total number of correct words from twenty five point five.
HINT and matrix tests use an adaptive method to find the fifty percent point, the SRT50.
Test yourself: On a QuickSIN list, the patient repeated sixteen key words correctly in total. What is the SNR loss?
The correct answer is B: nine point five decibels, a moderate SNR loss.
Twenty five point five minus sixteen is nine point five; this value falls between seven and fifteen.
In the next lesson, we will see when masking is needed in speech tests and how to calculate the noise level.
Subscribe to the channel so you never miss a new lesson. See you soon!

Keywords

Click a keyword to search all lessons for that term.

Prerequisites

Comments

Questions and comments appear after approval.

No comments yet. Be the first to comment.