PSYCHOMETRIC VALIDATION OF SPEECH PERCEPTION IN NOISE TEST MATERIAL IN ODIA (SPINTO)

Transcription

1 PSYCHOMETRIC VALIDATION OF SPEECH PERCEPTION IN NOISE TEST MATERIAL IN ODIA (SPINTO) PURJEET HOTA Post graduate trainee in Audiology and Speech Language Pathology ALI YAVAR JUNG NATIONAL INSTITUTE FOR THE HEARING HANDICAPPED, EASTERN REGIONAL CENTRE B.T.ROAD, NIOH CAMPUS, BONHOOGHLY, Kolkata-90. *PALASH DUTTA Lecturer (speech and hearing) ALI YAVAR JUNG NATIONAL INSTITUTE FOR THE HEARING HANDICAPPED, EASTERN REGIONAL CENTRE B.T.ROAD, NIOH CAMPUS, BONHOOGHLY, Kolkata-90. INDRANIL CHATTERJEE Lecturer (speech and hearing) ALI YAVAR JUNG NATIONAL INSTITUTE FOR THE HEARING HANDICAPPED, EASTERN REGIONAL CENTRE B.T.ROAD, NIOH CAMPUS, BONHOOGHLY, Kolkata-90. ASHOK KUMAR SINHA Director ALI YAVAR JUNG NATIONAL INSTITUTE FOR THE HEARING HANDICAPPED, MUMBAI Abstract *Corresponding Author One of the major problems of the hearing impaired population is speech comprehension in noise. There are various tests to assess the speech perception in noise but currently Quick SIN and Hearing in noise test (HINT) are considered as standard measures of speech perception in noise. There is few standardized speech perception in noise tests,which uses sentences, are available in Indian languages but as per our knowledge there is no speech perception in noise test are available in odia language. Since language is an important issue for developing and standardization of a test battery. It has an important influence on the outcomes in test which are linguistically loaded. Therefore it s a need to develop a standardized test in odia to measure the speech perception ability in noise. 116

2 Introduction: To develop this test materials sentence selection was done and familiarity check was done. Then selected sentences were recorded by female talker in presence of multitalker babble using Nuendo software. The recorded material was presented using a CD player routed through an Audiometer. Scoring was done and SNR loss was calculated. Results showed that the p valued and for normal and hearing impaired group respectively in Two way ANOVA, suggested all 7 lists were equivalent. And the mean SNR-50 was 2.74 db in normal hearing population and the mean SNR loss for hearing impaired group was 9.09 db. Test retest reliability also showed higher test retest reliability. Therefore this can be used for evaluation of speech perception in noise in odia population. KEY WORD- speech perception, hearing impairment, normal hearing, multitalker babble The persons with hearing impairment often report to have problem in understanding the heard speech in presence of noise, though they are able to hear the message but are unable to understand the meaning, conveyed through the message. Therefore, an appropriate material is highly needed to test the speech perception ability of the hearing impaired persons. Although, there are a few standardized test materials available in English to test the speech perception ability with presence of back ground noise, but these tests cannot be administered to native Odia population due to the language variation. Thus there is an exceptional requirement to develop the test materials in the Odia language to assess the speech perception in presence of multi talker babble noise for the native Odia population. Need of the study: There are various tests to assess the speech perception in noise but currently Quick SIN and Hearing in noise test (HINT) are considered as standard measures of speech perception in noise. There is few standardized speech perception in noise tests, which uses sentences, are available in Indian languages but as per our knowledge there is no speech perception in noise test are available in odia language. Therefore it s a need to develop a standardized test in odia to measure the speech perception ability in noise. This test would be highly useful in selecting amplification devices for the hearing impaired individuals. METHODOLOGY The present study was to develop and standardize the speech perception in noise test material in Odia (SPINTO). This study was done in five phases with appropriate statistical analysis at every required phase i.e. Phase 1: The development of test material, Phase 2: Administration of SPINTO material, Phase 3: Test Retest reliability, Phase 4: Compilation of data. Phase 5: Statistical analysis Participants: In this study we have taken 2 groups, 30 person in each group, in the age group of years (Mean: years, SD: 5.7) has participated. Candidacy selection criteria: For normal hearing group: a) None of the participants had any Hearing Impairment. Their hearing thresholds were below 20 dbhl across the Audiometric frequencies. Mean pure tone threshold for right ear and left ear was dbhl (SD: 1.94) and 18 dbhl (SD: 1.54). b) Their first language (L 1 ) was Odia. c) None of the participants had any auditory processing disorder. d) None of the participants had any Neurological and Cognitive Impairment. e) None of the participants had any middle ear disorder For hearing impaired group: 117

3 a) They had mild to moderately severe sensorineural hearing loss. Mean Pure tone threshold for Right ear and Left ear was db (SD: 5.71) and db (SD: 5.24) respectively. b) None of the participants had any intellectual or cognitive impairment. c) None of the participants had any other neurological impairment. Procedure: Phase I: Development of test material Collection of word lists: The material used for the study was taken from different odia news paper and story books. Before the selection of sentences some decisions were made such as reasonable degree of homogeneity would be maintained in sentence length (5-7 word per sentence). Therefore only those sentences are selected which has fulfilled criteria Familiarity of the key words in the sentence: At first, 60 sentences (containing 300 key words) were collected and these sentences were given to five native adult Odia speakers, who had at least 15 years of formal education in Odia. They were instructed to underline the key words and also to rate their familiarity of the key words from all the sentences using a five point rating scale (1 being least common, 5 being most common, while 2, 3 and 4 representing the intermediate values) Selection of the sentences: Finally, 42 sentences were selected from this pool based on the participant s rating of key word familiarity. The criterion used was 100% agreement between all the 5 subjects. The Cronbach s alpha was calculated to be Using these 42 sentences which were distributed in 7 lists with 6 sentences in each list was prepared. Each sentence had five key words. Sentences were randomly divided into different lists Recording Parameters: These sentences were recorded by a native female Odia speaker in sound recording studio using professional voice recording and editing software called Nuendo (version 6.0). The speaker was asked to speak at best quality and instructed to maintain equal stress and loudness during recording of all sentences. The microphone was placed at a distance of 6 inches from the mouth. The signal was digitized at sampling rate Hz. To generate sentences with different SNRs, Multi talker babble was used. For this purpose, two minutes of news paper reading was recorded from eight native Odia speakers separately. Later each of these recordings was digitally mixed using Nuendo (version 6.0) Software. In each list the first sentence was a SNR, in which 100% speech recognition scores was present, and the SNR was be reduced in 5 db steps for the subsequent sentences till the last sentence which had a SNR, where near 0% speech identification scores was present. With these ranges of SNRs, it was intended to measure the SNR that resulted in 50% speech recognition scores. The signal- to- noise ratios was manipulated by varying the amplitude level of the babble while keeping the amplitude of the spoken sentences constant. This was done by using Nuendo (version 6.0) Software. It was ensured that the noise and speech signals were of equal loudness by normalizing the signals. Prior to each list a 1000 Hz calibration tone was recorded in each word list, and was used to adjust the VU meter of the audiometer to zero. Thus, the test was composed of lists of sentences embedded in multi talker babble Phase II: Administration of SPINTO material: At first pure tone thresholds were calculated for all subjects. The testing was done for the frequencies 250 Hz to 8000 Hz for air conduction and 250 Hz to 4000 Hz for bone conduction. All the subjects were tested for normal middle ear function using the tympanometery and acoustic reflex test. Duration of testing was thirty minute for each patient. All the testing was conducted in a double room Audiological testing suite. The subjects were tested using a MAICO MA 53 clinical audiometer using TDH-39 headphones and B 71 bone vibrator. The audiometer was calibrated according to IEC , 2012 standards. Immittance testing was done using GSI 39 Immittance Audiometer. The SPINTO test was conducted with participants seated in a comfortable position with adequate illumination. The material developed for the study was played through a windows based personal computer with CD ROM and connected to the audiometer by means of an audio cable. These lists were presented at 70 dbhl presentation level binaurally via head phone in a pseudo random order using MAICO MA 53 clinical audiometer in normal group. While the presentation level for Hearing Impaired group was at Loud but Ok 5 level 2. The presentation level was manipulated using the attenuator dial of the audiometer. The listener s task was to repeat each word of the sentence presented. Participant s response was recorded in Tape recorder for further analysis. 118

4 Instruction: The participants of both groups were instructed before the test as Imagine that you are at a party. There will be a woman talking and several other talkers in the background. The woman s voice is easy to hear at first, because her voice is louder than the others. Repeat each sentence the woman says. The background talkers will gradually become louder, making it difficult to understand the woman s voice, but please guess and repeat as much of each sentence as possible Scoring: Five key words were scored in each sentence. The key words were underlined on the score sheets. One point was given for each key word repeated correctly. The number of correct words for each sentence was written in the space provided at the end of the sentence and the total correct calculated for the list. Average SNR 50 was calculated for normal group, based on the total score of the participants. SNR 50 and SNR Loss were calculated for each list by using the 4 procedure Phase III: Test- Retest Reliability. To look for there being no confounding factor during the intervening time interval, the subjects who were participated in phase II retested after one month interval by using the same recorded SPINTO material to check the test retest reliability Phase-IV: Compilation of data. All the collected data were compiled in Microsoft excel sheet for further statistical analysis. All data of the normal hearing group as well as hearing impaired group were assembled for test and retest condition Phase-V: Statistical analysis. The statistical analysis was done by using appropriate statistical method. Equivalency of test lists in normal hearing as well as in hearing impaired group was tested using Two way ANOVA. Two way ANOVA was also used to see the effect of SNR on speech identification scores in both groups. Comparison between normal and hearing impaired was done using Two way ANOVA. And the test retest reliability was also checked by Pearson s correlation and Two way ANOVA. For all the above statistical analyses Data Analysis pack of Microsoft Excel was used (version: Microsoft office, 2007). RESULTS The main goal of this study is to find out the speech identification score as well as to find out the Signal to noise ratio loss (SNR LOSS) for hearing impaired. To investigate the objectives of the present study, statistical analysis is done which is shown under following experiments, which shows the benefit of the test material for Odia population as well as reliability and validity of test material. Phase 1: Equivalency of the sentences. Phase 2: Calculation of Signal to noise ratio 50 (SNR 50). Phase 3: Calculation of Signal to noise ratio loss (SNR LOSS). Phase 4: Discriminant validity with respect to hearing loss and normal hearing. Phase 5: Tests retest reliability Phase 1: Equivalency of the Lists. In the Phase 1 of this study Two way ANOVA was applied to see the equivalency of the lists in Normal Hearing as well as in Hearing Impaired population. The results are depicted in below Tables. 119

5 Table-5.1: Two way ANOVA for List Equivalency in normal. Source of Variation SS Df MS F P-value F crit LISTS SNR Interaction Within Total The Table-5.1 depicted the result of Two way ANOVA for List equivalency. In the above table the p value is for lists which shows that there is no significant difference in list which means list are equivalent. For the Signal to Noise ratio (SNR), on which each sentence are made in the list, has p value is 0. This indicates that the identification of key words is significantly affected by the Signal to noise ratio. The interaction value is which shows that the SNR and lists are independent of each other. Table-5.2: One way ANOVA for difference in SNR in normal. Source of Variation SS Df MS F P-value F crit Between Groups Within Groups Total The Table-5.2 depicted the result of one way ANOVA. This is done to see the effect of SNR on identification scores of key words, the value of p is 0 which indicates that there is significant difference in scores due to the SNR variation on sentences. Table-5.3: Two way ANOVA for List equivalency in hearing impaired. Source of Variation SS Df MS F P-value F crit Lists SNR Interaction Within Total In the Table-5.3, the results of Two way ANOVA is shown. This is done to see the lists equivalency in Hearing Impaired. The p value is which indicates that there is no significant difference list that is the lists are equivalent in Hearing Impaired also. The p value for SNR is 0 which means there is significant difference in scores due to the SNR variation. The interaction value is which shows there is small effect of SNR as well as lists on scores but this is statistically not significant at 1% of significance level. So this could be negligible Phase 2: Calculation of Signal to noise ratio 50 (SNR 50). In the Phase 2 we have calculated the SNR 50 using 4 which is based on the Spearman-Karber equation. The formula, provided below, is used to derive the 50% correct point (T50%). 120

6 Average SNR Loss Mean SNR 50 T50% = I + ½(d)-d(r)/n. Whereas I represents the initial test presentation level, d the db decrement between presentation levels, r the number of correct responses overall and n the number of test items presented at each level. Therefore, the formula is simplified if the number of items per presentation level is the same as the decrement size. The results for SNR 50 for normal hearing are depicted in below Graphs. In this study the average SNR 50 for normal hearing was found 2.74 db. This value was 2 db in English language, when they used Female speaker 3. In the Figure 5.1, 7 different lists are shown in X axis and the mean SNR 50 for 30 normal hearing people on Y axis. We have also calculated the average SNR 50 across 7 lists which is 2.74 db. This value tells us that 2.74 db SNR is required to achieve 50% of speech identification score in noise Mean SNR Lists Figure 5.1: Mean SNR 50 for 30 Normal Hearing Group across 7 lists. Phase 3: Calculation of Signal to noise ratio loss (SNR Loss). In Phase 3 we have calculated the SNR Loss for normal hearing as well as for hearing impaired. In this study Tillman-Olsen method was used to calculate SNR Loss. Since SNR 50 in this study for normal hearing group is Therefore formula for SNR LOSS was, SNR Loss = SNR In this study the average SNR Loss for normal is dB which is very near to 0. And the Average SNR Loss for hearing impaired group is 9.09 db. The Average SNR Loss for normal as well as for hearing impaired is depicted in below Graphs Lists AVG SNR LOSS Figure 5.2: Average SNR Loss for Normal Hearing Group across 7 lists. 121

7 Mean SNR LOSS The Figure 5.2 depicted the average SNR Loss for Normal Hearing group (30 subjects) across the 7 lists. In which the x axis represents the number of lists and the y axis represents the mean SNR Loss of the normal hearing group. The above figure shows average SNR Loss values are very near to 0. This also shows the homogeneity of the sentences and lists. The same procedure was applied to calculate SNR Loss for Hearing Impaired group, which shows significant higher SNR Loss than Normal Hearing group. The following graph depicted the Average SNR Loss for Hearing Impaired Group Mean SNR Loss Lists Figure 5.3: Mean SNR Loss for Hearing Impaired group across 7 Lists. The Figure 5.3 depicted the average SNR Loss for Hearing Impaired group across the 7 lists. In the above figure X axis represents number of lists and Y axis represents mean SNR Loss. The Mean SNR Loss for Hearing Impaired Group is 9.09 db. This is significantly higher than the normal SNR Loss value. This also shows that a hearing impaired required more SNR than a normal hearing person Phase 4: Discriminant validity with respect to hearing loss and normal hearing. Table-5.4: Two way ANOVA for Discriminant validity with respect to Hearing Impaired and Normal Hearing. Source of Variation SS Df MS F P-value F crit Lists Normal Vs. HI E Interaction Within Total The Table 5.4 depicted the result of Two way ANOVA for Discriminant validity with respect to Hearing Impaired and Normal Hearing. The p value for Discriminant validity is 4.4E-255 which is near to zero suggestive of significant differences between normal hearing and hearing impaired even if we allowed for variation through lists. So these results show that, this test can differentiate between hearing impaired and normal hearing. P value for Lists is shows that there is no significant variation in lists. P value for Interaction is suggests that the list and Normal and hearing impaired values are independent Phase 5: Tests retest reliability. This test was administered to 60 participants (30 normal hearing and 30 hearing impaired). The same test was administered on the 60 participants of the original sample for the second time one month from the first time of measures. The below tables show the results of Two way ANOVA for test retest reliability. 122

8 Table- 5.5: Two way ANOVA for test retest reliability in normal. Source of Variation SS Df MS F P-value F crit Lists E Test vs. Retest Interaction Within Total The Table 5.5 depicted the result of Two way ANOVA test for Test Retest reliability in Normal Hearing Population. The p value of Test Retest reliability is which shows no significant difference in results of Test Retest values. This is suggestive of the high Test Retest reliability. The p value for list is 1.68E-05 which is near to 0 suggested that there is inter list SNR Loss variation present. The interaction value is which shows the list and test retest are independent of each other. Table-5.6: Two way ANOVA for test retest reliability in Hearing Impaired. Source of Variation SS Df MS F P-value F crit Lists Test vs. Retest Interaction Within Total Table 5.6 depicted the results of Two way ANOVA for Test Retest reliability in Hearing Impaired. The p value for Test Retest reliability is which shows that the test is highly reliable and there is no significant difference in the results of initial test and 30 days later test. The p value for list is which suggests that there is inter list SNR Loss variation present. The interaction value is shows the list and test retest values are independent to each other. To see the test retest reliability we have also applied Pearson s correlation in both normal hearing and hearing impaired groups. The test data and retest data were used to see the correlation. The following table shows the correlation of test retest values of normal and hearing impaired group. Table-5.7: Correlation of test retest values for normal hearing group. Test Retest Test 1 Retest ** 1 ** Correlation. The Table 5.7 depicted the Pearson s correlation of test and retest data of normal hearing group. The correlation value was which suggests that there is high correlation between test and retest data in normal hearing group. 123

9 Table-5.8: Correlation of test retest values for hearing impaired group. Test Retest Test 1 Retest ** 1 ** Correlation. The Table 5.8 depicted the Pearson s correlation of test retest data of hearing impaired. The correlation is which suggest the test and retest data are highly correlated for hearing impaired group. DISCUSSION Speech perception includes the process of hearing speech sounds, their interpretation and understanding. But one of the basic problems in the study of speech is how to deal with the noise in the speech signal. Since our daily life is not completely quite and there is always some background noise during conversation. Therefore it is essential to assess the speech perception in noise. The present study was done taking in account of equivalency of the list as well as signal to noise ratio for normal and hearing impaired group. In this study SNR 50 and SNR Loss was calculated using the Tillman Olsen method 4. The following headings were discussed below: 6.1 Development of test material. 6.2 Equivalency of the test lists Effect of SNR Calculation of SNR Calculation of SNR Loss. 6.6 Discriminant validity with respect to hearing loss and normal hearing. 6.7 Test Retest reliability Development of test material. During collection of sentences the major focus was given on the sentence length. So that the length of the sentences will remain same across the list and there will be no confusion about the key words present on the sentences. However, sentence length was kept 5-7 words. The sentences were selected from Odia story books and News Paper. Total 60 sentences (containing 300 key words) were selected. The second most important thing is selection of the key words and judgment of familiarity. To test the familiarity of key words, these 300 key words were given to 5 Native odia speakers, who had at least 15 years of formal education in odia. They were instructed to underline the key words and also to rate their familiarity of the key words from all the sentences using a five point rating scale (1 being least common, 5 being most common, while 2, 3 and 4 representing the intermediate values). When each key word was rated a minimum of 4 or more than that, the words have been then defined as familiar. Out of 60 sentences, the 42 sentences thus formed contained 210 words each based on familiarity weightage. The Cronbach s alpha was calculated to be These sentences were then randomly distributed in to 7 lists, where each lists contains 6 sentences. 124

10 Mean Scores The third most important aspect of the material is the background noise with different SNRs. For this purpose multitalker babble was used. To create multitalker, two minutes of news paper reading was recorded from eight native Odia speakers separately. Then these recordings were digitally mixed using Nuendo (version 6.0) Software. The total 6 SNR were selected for each list in which the sentences were recorded. The first sentence was with an SNR in which 100% speech recognition score was present, and the SNR was reduced in step of 5 db for the subsequent sentences. The last sentence had a SNR in which the speech identification score was 0%. Based on these, total 6 SNR were selected, which were recorded in descending order in the lists. In this study the range of SNR were 25 db, 20 db, 15 db, 10 db, 5 db and 0 db. Killion et al., 2004 also selected these range of SNR for their Quick SIN test. In which the first sentence was with 25 db SNR and the SNR reduced by 5 db steps in subsequent sentences and the last sentence was with 0 db SNR. But these SNR values can vary with language to language such as, in the study 1 developed a quick SIN test in Kannada language in which they have selected different range of SNR which are 20 db, 15 db, 10 db, 5 db, 0 db, -5 db and -10 db Equivalency of the test lists. Two way ANOVA was applied to see the equivalency of the test and the effect of SNR on speech identification score. The p value is for lists this shows that there is no significant difference in list which means list are equivalent. In the present study mean speech identification scores for each list were calculated. The standard deviation was also calculated. The mean standard deviation across sentences for all normal hearing subjects was 0.43 db and the mean standard error was 0.08 db. The small standard deviation across the subjects also shows a good equivalency of the lists. But study 3 found the standard deviation was 1.25 db and the standard error was 0.25 db. This difference is may be due to they have calculated single list standard deviation and standard error. The Figure 6.1 depicted the mean correct scores for each list and standard deviation. The maximum possible correct score was 30. The X axis represented the Sentence lists and Y axis represented Mean scores in the below figure MEAN SCORE Sentence Lists Figure 6.1: Mean speech identification scores of Normal Hearing Group for each list. Error bars indicate 1 SD of error. 125

11 Mean scores The equivalency of the lists was also checked for the Hearing Impaired group. As the ANOVA results shows the p value is Thus there is no significant difference for the lists. This means all the lists has good equivalency for hearing impaired group. The results also can also be seen through graph which shows that there is small Standard deviation, which also represents the equivalency of the lists. The mean standard deviation across the list for all hearing impaired subjects was 1.58 db and the standard error was 0.29 db. These values are also supported 3. They reported that their standard deviation was 1.4 db across the lists which is not significantly different from this study. They have also reported this value was 1.9 db in SIN test. The below figure shows the mean scores of the all subjects across lists and the standard deviation. In figure 6.3, the X axis represented the sentence lists and the Y axis represented the mean scores Mean Scores Sentence Lists Figure 6.2: Mean speech identification scores of Hearing Impaired Group for each list. Error bars indicate 1 SD of error Effect of SNR. In this study, Two way ANOVA was applied to assess the effect of signal to noise ratio (SNR) on speech identification scores for normal hearing group. The p value was 0, this suggests there is significant difference in scores across SNRs. To reconfirm One way ANOVA was applied for six SNR levels. Which shows again the p value is 0. This suggests the identification scores varies with SNR levels. This is because when the SNR is high the speech identification was better and as the SNR reduces the identification scores reduces. Therefore the higher SNR, higher the score and the lower the SNR, lower the scores. The effect of SNR on speech identification scores was also evaluated for hearing impaired group. The Two way ANOVA was applied to see the effect. The p value was found 0 which is again significant as in normal hearing group. So the effect of SNR on 126

12 Mean SNR 50 speech identification scores is also present in the hearing impaired group. On account of the masking, the external redundancy present in the speech signal decreases and it was difficult for the subjects to perceive the signal Calculation of SNR 50. In this study SNR 50 was calculated using Tillman-Olsen method 4. This method is based on the Spearman-Karber equation by Kinney 3. The formula, provided below, is used to derive the 50% correct point (T50%). T50% = I + ½(d)-d(r)/n. Whereas I represents the initial test presentation level, d the db decrement between presentation levels, r the number of correct responses overall and n the number of test items presented at each level. Therefore, the formula is simplified if the number of items per presentation level is the same as the decrement size. The number of test items per level for this test is the same as decrement size, which is 5. Each key word can be thought of as representing 1 db. First SNR 50 was determined by adding one half step size or 2.5 db to the initial level of presentation of SNR that is 25 db. Therefore it would be = Then SNR 50 was calculated by subtracting total correct identification scores from Most speech recognition score provides a percentage of correct score or SNR required for 50% correct point. This test allows the opportunity to compare to normal performance. Therefore the resultant value provides an individual s SNR Loss. In this study the average SNR 50 for normal hearing was found 2.74 db. This value was 2 db in English language, when they used Female speaker 3. This value was better when they used male speaker that is -3 db. A Study done in Kannada language 1 shows this value was db when they used male speaker. However, in this study the SNR 50 is 2.74 db which is comparable to 2 db, since this study has female speaker. The below figure depicted the mean SNR 50 across all normal hearing subjects. In below figure X axis represented the number of subjects and the Y axis represented the Mean SNR Mean SNR Subjects Mean SNR 50 Figure 6.3: Mean SNR 50 for Normal Hearing Group. The same procedure was also applied for calculation of SNR 50 for hearing impaired population. The result shows that the mean SNR 50 for hearing impaired group was db. This shows that hearing impaired population required more SNR for 50% of the speech identification score. 127

13 Mean SNR Loss 6.5. Calculation of SNR Loss. SNR Loss is the db increase in signal-to-noise ratio required by a hearing-impaired person to understand speech in noise, compared to someone with normal hearing. In this study Tillman-Olsen method was used to calculate SNR Loss. Since SNR 50 for normal hearing person is 2.74 in this study. Therefore formula for SNR Loss will be, This can be simplified as, SNR Loss= SNR SNR Loss= (27.5-Correct scores) =24.76-Correct scores. In this study the average SNR Loss for normal is dB which is very near to 0. This also represents the homogeneity of the lists. The figure 6.4 depicted the mean SNR 50 across 30 normal hearing subjects Subjects MEAN SNR LOSS OF NORMAL GROUP Figure 6.4: Mean SNR Loss of Normal Hearing group. This is also supported by a study by Avinash and Mati 1. They also reported that there SNR 50 for the normal hearing population was around 0. The Average SNR Loss for hearing impaired group was also calculated using the same formula. This shows the mean SNR Loss for hearing impaired group was 9.09 db. This is also supported by the Killion et al. 3. They show the SNR loss for hearing impaired varies from 2 to >15 db. This variation may be due to the type and degree of hearing loss. In this study wide range of degree of hearing loss was included. Therefore the value is 9.09dB Discriminant validity with respect to hearing loss and normal hearing. In this study to see the Discriminant validity with respect to Hearing Impaired and Normal Hearing, Two way ANOVA was used. The p value for Discriminant validity is 4.4E-255 which is near to zero suggestive of significant differences between normal hearing and hearing impaired even if we allowed for variation through lists. So these findings suggest that, this test can be used to differentiate between hearing impaired and normal hearing based on SNR Loss. P value for Lists is shows that there is no significant variation in lists. P value for Interaction is suggests that the list and Normal and hearing impaired values are independent. Below figure shows the mean SNR Loss difference in normal hearing population as well as in hearing impaired population. There is more variation in the SNR Loss in hearing impaired group due to wide range of degree of hearing loss was included in this study. 128

14 Mean SNR Loss MEAN SNR LOSS OF NORMAL GROUP MEAN SNR LOSS OF HI GROUP Subjects Figure 6.5: Mean SNR Loss of normal and hearing impaired groups. 6.7.Test Retest reliability. Reliability and the consistency of a test or measurement, is frequently quantified in the movement sciences literature. In this study to see the test retest reliability this test was administered on the 60 participants of the original sample (30 normal hearing and 30 hearing impaired) for the second time one month from the first time of measures. Two way ANOVA was used for the statistical analysis in Normal Hearing Population to see if there is any significant variation in test and retest scores. The p value of Test Retest reliability is which shows no significant difference in results of Test Retest values. This is suggestive of the high Test Retest reliability. The p value for list is 1.68E-05 which is near to 0 suggested that there is inter list SNR Loss variation present. The interaction value is which shows the list and test retest are independent of each other. Correlation statistics to check test retest reliability in normal hearing and hearing impaired group. This show the correlation was in normal hearing group which suggest there is significant correlation between test and retest data. Therefore the test is highly reliable for normal hearing population. The correlation was also calculated for hearing impaired group which show the correlation value was This suggests there is significant correlation between test retests values. So this test is highly reliable for hearing impaired population. REFERENCE 1.Avinash, M. C., Meti, R., & Kumar, U. (2010). Development of Sentences for Quick Speech innoise (Quick SIN) Test in Kannada. Journal of Indian Speech and Hearing Association, 24(1), Etymotic Research (2001). Quick SIN Speech in Noise Test Version 1.3.Elk Grove Village, IL. Feng, Y., Yin, S., Kiefte, M., & Wang, J. (2010). Temporal Resolution in Regions of Normal Hearing and Speech Perception in Noise for Adults with Sloping High Frequency Hearing Loss, Journal of Acoustical Society of America, 112, Killion, M. C., Niquette, P. A., Gudmundsen, G. I., Revit, L. J., & Banerjee, S. (2004). Development of a quick speech-in-noise test for measuring signal to noise ratio loss in normal hearing and hearing impaired listeners. Journal of the Acoustic Society of America, 116(4), Tillman, T. W., & Olsen, W. O. (1973). Speech audiometry. Modern developments in Audiology, 2, Valente, M., & Van Vliet, D. (1997). The independent hearing aid fitting forum (IHAFF) protocol. Trends in Amplification, 2(1),