|Year : 2018 | Volume
| Issue : 2 | Page : 109-113
Measurement of speech in noise abilities in laboratory and real-world noise
Bhanu Shukla1, B Srinivasa Rao2, Udit Saxena3, Himanshu Verma4
1 Assistant Professor, RP Institute of Speech & Audiology, Karnal, Haryana, India
2 Assistant Professor, AYJNISHD (SRC), Secunderabad, India
3 Reader, RP Institute of Speech & Audiology, Karnal, Haryana, India
4 Assistant Professor, Ashtavakra Institute of Rehabilitation Sciences & Research, New Delhi, India
|Date of Web Publication||4-Sep-2018|
Prof. Himanshu Verma
H. No - 19A, H-Block, Street-17, Gulab Farm, Hamdard Road, Sangam Vihar, New Delhi - 110 062
Source of Support: None, Conflict of Interest: None
Aim: The present study aimed to investigate speech in noise perception abilities in normal-hearing adults using different types of noise (i.e., speech babble, traffic noise, and speech spectrum noise) and at different signal-to-noise ratios (SNRs) (i.e., +5 dB, 0 dB, −5 dB, and −10 dB). Methods: A total of 109 individuals with mean age of 23 years were participated in the study. All participants had English as second language and Telugu as native and first language. English disyllabic words (Hrish et al.) were used as speech stimuli under different noises and different SNRs. Results: Results have shown that for any type of noise the speech perception scores changed with the variance in SNRs, the scores have decreased when the SNRs were decreased from 0 dB to −10 dB SNR, and the scores have increased when SNRs increased from 0 dB to +5 dB SNR. Conclusion: From this study, it can be concluded that speech perception score in noise depends on the type of noise used in testing the speech in noise abilities. This factor is very important in the selection of the noise type when measuring speech perception in the presence of noise. It was also seen that change in noise level also had a different impact on speech perception in noise abilities.
Keywords: Disyllabic words, speech babble, speech perception, speech spectrum noise
|How to cite this article:|
Shukla B, Rao B S, Saxena U, Verma H. Measurement of speech in noise abilities in laboratory and real-world noise. Indian J Otol 2018;24:109-13
| Introduction|| |
Speech in noise tests
Speech perception abilities get adversely affected in the presence of noise. Carhart and Tillman  highlighted the importance of estimating speech perception in noise in the regular auditory diagnostic battery. Several speech-in-noise (SIN) tests have been developed since then. Connected sentence test, hearing in noise test (HINT), words in noise (WIN), QuickSIN test (QuickSIN), Bamford-Kowal-Bench SIN Test (BKB-SIN), and listening in spatialized noise-sentences (LiSN-S) are some of them.
The HINT was developed by Nilsson et al. for the measurement of reception threshold for sentences in quiet and in the presence of noise. The goal of the HINT is to provide a reliable and efficient tool to estimate hearing handicap, directional hearing, hearing-aid benefits, and hearing-aid comparison., The HINT includes 25 phonemically balanced lists of 10 sentences which were adapted from the BKB sentences (Bench and Bamford, 1979 as cited in Nilsson et al., 1994) which were designed to be suitable for age from 4.6 years (Kowal, 1979). The BKB sentences included in the HINT were revised to remove British idioms and to equate the length of the sentences.
The QuickSIN test , and the BKB-SIN test  were developed to measure the signal-to-noise ratio (SNR) loss in decibels. SNR loss is described as the dB increase in SNR required by an individual with hearing impairment to understand speech in presence noise, compared to someone with normal hearing. The QuickSIN is a quick and improved version of the SIN Test developed by Etymotic Research (1993 as cited in Killion et al., 2004). The goals of the QuickSIN test are to provide a fast means of measuring SNR loss, quantify the benefits of directional microphones, and help the audiologist in choosing appropriate amplification options for individuals with hearing loss. The QuickSIN is comprised 12 lists (female talker) each containing six sentences with six additional lists (three lists for practice and another three for research). Each sentence consists of five keywords and each keyword is worth one point. Sentences used in the QuickSIN were adapted from Institute of Electrical and Electronics Engineers (IEEE) sentences which were later equalized to account for the high-frequency attenuation present in the original recording from Massachusetts Institute of Technology (MIT; Fikret-Pasa, 1993 as cited in Duncan and Aarts, 2006). These sentences provide limited contextual cues to the listeners. The BKB-SIN was developed to overcome the shortcoming of the QuickSIN Test in testing young children that includes the use IEEE sentences in the QuickSIN which are approximately at high school language level. In addition, these sentences are lengthy which causes difficulties in testing cochlear implant users and adults with auditory memory deficits. The BKB-SIN uses the BKB sentences. The BKB-SIN consisted of 18 lists pair. 1–8 lists pair has 10 sentences in each list and 9–18 lists pair has eight sentences in each list. First sentence of each list has four keywords and rest have three keywords. Only 1–8 lists are recommended for testing normal-hearing listeners.
The LiSN-S test  is a newer version of the original listening in spatialized noise–continuous discourse test  and was developed to assess SIN perceptual abilities in children as young as 5 years by incorporating a simplified and more objective response protocol. Target sentences (120 sentences; Female talker) used in LiSN-S were written by Australian Speech Language Pathologist. These sentences were constructed to be suitable for children from 4.6 years. Distracter stories are used as noise in LiSN-S. The stories were recorded by three female talkers (including the one who recorded target sentences). Both target and distracters are synthesized with head related transfer function (HRTF). Using HRTFs, the target sentences are perceived to be coming from 0° azimuth whereas the distracter stories (masker) vary according to the spatial location (0°, +90° and −90° azimuth), the vocal identity of the speaker(s) of the stories (same as, or different from, the speaker of the target sentences), or both. This test configuration results in four distracter conditions: (1) same voice at 0° (SV0°); (2) same voice at +90° (SV +90°); (3) different voices at 0° (DV0°); and (4) different voices at +90° (DV +90°). Performance on the LiSN-S is evaluated in terms of low- and high-cue SRT and also on three “advantage” measures. These advantage measures are the benefit in dB gained by cues-like different talker (pitch), spatial cues, and both talker and spatial cues compared to the low-cue SRT condition where no cues are present.
The WIN was developed by Wilson, to measure the ability to understand speech in multitalker babble. WIN is different from HINT, QuickSIN and LISN-S because it uses monosyllabic words as speech materials while others use sentences. Wilson  advocated for monosyllable words as speech material over sentences for SIN tests because: (1) sentence materials are not used widely in clinical audiology practice; (2) repeating sentence materials, especially in the presence of noise, involves many issues other than speech recognition tasks such as recognition versus recall, memory, recency and primacy effects, bottom-up versus top-down information processing, and the multiplicative effects of various degradations on the recognition of the speech signal. In addition, the use of monosyllabic words makes the SIN measure free of linguistic contextual cues. On the other hand, sentences spoken with natural dynamics are advocated to be used as test speech material, for the reason that it provides a real-world representation and has a dynamic range that is more natural than the monosyllabic words.
Measurement procedures adopted in speech in noise tests
The HINT uses an adaptive procedure to converge at the 50% performance point for both in quiet and in noise measurements. The HINT follows a 100% criterion to be considered as the correct response for a sentence with some flexibility given in responses for articles. Like the HINT, the LiSN-S is an adaptive test. For each condition, SNR is adapted by increasing or decreasing the target level in 4 dB steps until the first reversal and then in 2 dB step to determine individual SRT in noise (dB SNR). SNR is decreased by 2 dB if a listener repeats more than 50% of the words correctly and is increased by 2 dB if listener scored <50%; however, SNR is not changed if a response of exactly 50% correct occurs. Testing stops itself when the listener either; (a) completes the entire 30 sentences in the list or (b) completes the practice sentences with a minimum of 17 target sentences, and their standard error (SE; which is calculated automatically in real time over the scored sentences) is <1 dB and the dB SNR appears on the screen.
The QuickSIN, the BKB-SIN, and the WIN tests use a descending level paradigm (method of constants) to calculate the correct score. This is followed by calculating 50% correct performance using the Spearman Karber. The Spearman-Karber formula  is 50% correct performance = i + ½ (d) – (d) (correct score)/w. In which i = the initial SNR, d = step size of presentation level, and w = the number of items at each level.
Limitations of presently available speech in noise tests and rationale for the present study
Speech materials, type of noise, and measurement procedures are some of the factors that may affect the SIN outcomes. Speech stimuli used in the test can be phonemes, words, or sentences which have different language demands that can influence test results. All the standardized tests presently available to measure speech in noise acuity are available only in English. Therefore, there is a high need to develop comparative speech perception in noise data in Indian languages.
Different noises (broadband, narrowband, speech-noise, speech-babble, etc.) produce different speech intelligibility scores even for similar SNR (Carhart, Tillman, and Greetis, 1969). The HINT uses a steady-state speech-shaped noise for the in noise measurements. This noise is spectrally matched to the long-term average spectrum of the target sentence stimuli. The HINT uses speech-shaped noise because of the stability of level of this noise which can increase the reliability of individual SNR scores. The QuickSIN, BKB-SIN, and WIN uses multi-talker babble (one male and three female voices) as noise for the SIN measurements. The reason behind using multi-talker babble is that it's a more real-world representation of background noise. Distracter stories are used as noise in LiSN-S. However, audiologist around the globe is unsure about the type of noise they should be using in measuring speech in noise abilities. It has been seen that most of the time audiologist go with the type noise easily available to them without any rationale.
As explained earlier most of the clinically used speech in noise tests measure dB SNR (signal to noise ratio) at 50% performance criteria. While administering HINT, LiSN-S, QuickSIN, BKB-SIN or WIN, it is either speech or noise that is kept constant while the intensity level of the other changes. In the real world environment intensity level of both noise and signal change with time. Therefore measuring of correct percentage scores will be more informative instead of measuring values in dB SNR. Further, especially in Indian languages, there is a lack of comparative data of speech in noise perception.
The present study will investigate speech in noise perception abilities in normal-hearing adults using different types of noise and at different SNRs.
- To estimate speech in noise in native Telugu normal-hearing listeners
- To compare speech in noise abilities in different types of noise (i.e., speech spectrum noise, multitalker babble, traffic noise)
- To compare speech in noise abilities at different SNRs (i.e., 5 dB SNR, 0 dB SNR, −5 dB SNR and −10 dB SNR).
| Methods|| |
A total of 109 subjects with mean age of 23 years were participated in the study. All participants had English as second language and Telugu as a first language. Among 109 subjects, 58 were male with mean age of 23.08 year and 51 were female with mean age of 22.98 year participants with any sensory acuity, neurological, behavioral, and psychological issues were excluded from the study.
English disyllabic word list (Hrish et al.) was used as the speech stimuli. It contains 20 words. Speech spectrum noise, multitalker babble, and environmental noise were used. Speech spectrum noise was generated using Praat software (which is freely available online software), multi-talker babble was recorded while simultaneously reading newspaper and traffic noise was recorded for real-world situation using Sony digital voice recorder.
All the measurements were carried out in a sound-treated two room situation. Ambient noise levels in the test rooms were as per the standards of ANSI S3.1 (1999).
All the testing was carried out in sound-treated double room setup. Participants were seated in the center of experimental setup at a distance of 1 meter from loudspeaker. Speech intelligibility scores (in percentage) were measured in quiet condition and in the presence of noise. In noise measurement was done at three SNRs i.e., 0, −5, and −10 dB SNR. Participants were instructed to listen to the speech stimulus while ignoring noise in noise condition and repeat whatever heard.
| Results and Discussion|| |
Paired t-test was computed using SPSS 20.00 version software (IBM corp.). t-test was used to see the effect of different SNRs using different types of noise (i.e speech spectrum noise, multitalker babble & environmental noise) on speech perception. In the following description, results are presented separately for speech babble, traffic noise, and speech spectrum noise.
Speech perception in presence of speech babble
As shown in [Figure 1] under speech babble noise participants scored highest speech perception score under +5 dB SNR condition whereas least score was obtained under −10 dB SNR condition.
|Figure 1: Mean and standard deviation of the speech perception scores obtained in speech babble at −10, −5, 0, and +5 dB signal-to-noise ratio|
Click here to view
Speech perception in presence of traffic noise
As shown in [Figure 2], participants scored highest speech perception score under 5 dB SNR condition using traffic noise whereas least was scored under −10 dB SNR condition.
|Figure 2: Mean and standard deviation of the speech perception scores obtained in traffic noise at −10, −5, 0 and +5 dB signal-to-noise ratio|
Click here to view
Speech perception in the presence of speech spectrum noise
As shown in [Figure 3], subjects scored highest speech perception score at +5 dB SNR (99.5) which is followed by 0 dB SNR (96.16) using speech spectrum noise whereas least was obtained at −10 dB SNR condition.
|Figure 3: Mean and standard deviation of the speech perception scores obtained in speech spectrum noise at −10, −5, 0 and +5 dB signal-to-noise ratio|
Click here to view
Speech perception in the presence of different noises and different signal-to-noise ratios
As shown in [Figure 4], the majority of participants scored highest at +5 dB SNR under all 3 different noise condition however as seen from [Figure 4], participants at +5 dB SNR and 0 dB SNR scored almost equal under speech spectrum noise.
|Figure 4: Mean and standard deviation of the speech perception scores obtained in three types of noise (speech babble, traffic noise, and speech spectrum noise) at different signal-to-noise ratios (−10, −5, 0 and +5 dB signal-to-noise ratio)|
Click here to view
| Discussion|| |
It is a well-known fact that speech perception gets adversely affected in the presence of noise. Carhart and Tillman  highlighted the importance of estimating speech perception in noise in the auditory diagnostic battery, but still these measures are not being generalized in the usual clinical measurements.
For any individual listening abilities in the presence of noise are much more important when compared to listening in quiet condition, as in present-day life, most of the times communication takes place in the presence of one or other type of noises, for example, domestic noises (air conditioners, refrigerators, coolers, television, etc.), industrial noises (machinery), cafeteria noise, traffic noise, etc., whereas audiological tests are done in sound-treated room where these real-world noises are restricted. Sometimes, noises such as speech spectrum noise, white noise, narrow band noise, etc., are used in certain audiological measurements; however, these noises are very different from the real-world noises.
Most of the tests that are used clinically to estimate speech in noise abilities, for example, HINT, Quick-SIN, BKB-SIN, LiSN, etc., these tests provide measure of speech in noise abilities regarding dB SNRs that informs us about the level of speech required above or below the noise level to give a certain speech perception score.
The need for the study arises from the fact that there is no environment in the present world that is abundant from the noise. However, the very fact is being ignored or went unnoticed in the field of audiology. Thus, the result of this study has in fact highlighted the necessity of involvement of noise as an additional variable in testing any auditory acuity. It was observed in literature that various studies have pointed out the fact that various types of noises do affect speech perception in different ways; however, the quantitative estimation was missing especially in the Indian population.
From the result of present study, it can be observed that % correct responses are at −10 dB, −5 dB, 0 dB, and +5 dB SNR, under speech babble are 48, 60.33, 81, and 93.5, under traffic noise are 55, 66.33, 88.66, and 98.83, and under speech spectrum are 69, 83.33, 96.16, and 99.5, respectively.
Above results have shown that for any type of noise the speech perception scores changed with the variance in SNRs, the scores have decreased when the SNRs were decreased from 0 dB to −10 dB SNR and the scores have increased when SNRs increased from 0 dB to +5 dB SNR. The presence of better SNR caused less pressure on the auditory system to encode the auditory information coded in the sentences presented as the stimuli and had to repeated back, as a task.
The result of the present study correlates with this study, as it was found that speech perception scores were worse at −5 and −10 dB SNRs in speech babble when compared to the traffic noise and speech spectrum noise and at 0 and +5 dB SNRs the scores for speech babble and traffic noise was not 100%, but for speech spectrum the scores at these SNRs were nearly 100%.
Further above, it is also shown that different noises affect the speech perception differently at different SNRs, Maximum percentage correct response score was obtained for traffic noise i.e., 98.83% at +5 dB SNR, and speech spectrum noise i.e., 99.5% at +5 dB SNR and minimum scores were obtained for speech babble i.e., 48% at −10 dB SNR, and for traffic noise, i.e., 55% at −10 dB SNR.
The effect of speech babble was most adverse when compared to traffic noise and speech spectrum noise. The reason behind this could be because of the presence of linguistic information because speech babble contains informative meaning full linguistic components which can interfere with the intelligibility of actual stimuli. There was also a significant difference in the scores of traffic noise and speech spectrum. The scores of traffic noise were worse than the speech spectrum noise this could be because of intensity fluctuations in the traffic noise. It is known that speech spectrum noise is a regularly used noise in audiology laboratory and have very low-intensity fluctuation over time.
| Conclusion|| |
From this study, it can be concluded that speech perception score in noise depends on the type of noise we are using in testing the speech in noise abilities. This factor is very important in the selection of the noise type when measuring speech perception in the presence of noise. It was also seen that change in noise level also had a different impact on speech perception in noise abilities. Therefore, the intensity level of noise exposure of the person should also be considered while evaluating speech in noise perception.
Noise is the main factor that can influence any auditory performance; however, the review of literature shows that its effect is rarely discussed or investigated. This study is an attempt to know the effect of various types of noise on perception of speech, at basic level. Further, there is a need for more number of studies to expand our knowledge in this area and add this component to the regular audiological evaluations in clinical practice.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| References|| |
Carhart R, Tillman TW. Interaction of competing speech signals with hearing losses. Arch Otolaryngol 1970;91:273-9.
Cox RM, Alexander GC, Gilmore C. Development of the connected speech test (CST). Ear Hear 1987;8:119S-26S.
Nilsson M, Soli SD, Sullivan JA. Development of the hearing in noise test for the measurement of speech reception thresholds in quiet and in noise. J Acoust Soc Am 1994;95:1085-99.
Wilson RH. Development of a speech-in-multitalker-babble paradigm to assess word-recognition performance. J Am Acad Audiol 2003;14:453-70.
Killion MC, Niquette PA, Gudmundsen GI, Revit LJ, Banerjee S. Development of a quick speech-in-noise test for measuring signal-to-noise ratio loss in normal-hearing and hearing-impaired listeners. J Acoust Soc Am 2004;116:2395-405.
Cameron S, Dillon H. Development of the listening in spatialized noise-sentences test (LISN-S). Ear Hear 2007;28:196-211.
House Ear Institute. Hearing in Noise Test operator's manual. Los Angeles, CA: Starkey Laboratories; 1995.
Duncan KR, Aarts NL. A comparison of the HINT and QuickSIN tests. J Speech-Lang Pathol Audiol 2006;30:86-94.
Cameron S, Dillon H, Newall P. Development and evaluation of the listening in spatialized noise test. Ear Hear 2006;27:30-42.
Wilson RH, McArdle RA, Smith SL. An evaluation of the BKB-SIN, HINT, QuickSIN, and WIN materials on listeners with normal hearing and listeners with hearing loss. J Speech Lang Hear Res 2007;50:844-56.
Hall JW 3rd
, Grose JH, Buss E, Dev MB. Spondee recognition in a two-talker masker and a speech-shaped noise masker in adults and children. Ear Hear 2002;23:159-65.
[Figure 1], [Figure 2], [Figure 3], [Figure 4]