Hearing Impairment

EPG Analysis

EPG data from 6 Greek speakers with profound hearing impairment have been recorded and analysed. The speakers were young adults or adolescents and the intelligibility of their speech varied from highly intelligible to unintelligible speech.
EPG patterns for several lingual consonants produced by the Greek speakers with hearing impairment are shown below.

In the phonetics laboratory we use EPG as a visual feedback technique for the teaching of lingual consonants to speakers with hearing impairment.

Acoustic Analysis

The following PhD research is being conducted by Anna Sfakianaki under Dr. Katerina Nicolaidis' supervision.

This study primarily attempts to investigate vowel-to-vowel carryover and anticipatory coarticulation effects in the speech of Greek adults with normal hearing and hearing impairment.

The hearing-impaired (HI) subjects are members of the Association of the Hard of Hearing of the county of Thessaloniki. Ten adults (balanced numbers for sex), 20-26 years old, with varying degrees of hearing impairment took part in the experiment. They can be categorised into two groups: Group A: subjects with a hearing loss of more than 100dB (profoundly deaf) and Group B: subjects with a hearing loss less than 100dB (75-99dB, severely deaf). Five adults with normal hearing (NH) constitute the control group. They were all born and raised in Thessaloniki, speak standard Greek with no detectable accent and are matched for sex and age with the hearing impaired subjects.

In order to investigate the relationship between coarticulation and speech intelligibility, an intelligibility test was set up. The HI subjects uttered 101 words and 25 sentences (8-14 syllables in total) which contain frequently used Greek clusters and all Greek phonemes in word-initial position at least once. An experiment with 30 naive listeners who judge the material is in progress. Finally an intelligibility score (1-5) will be given to each HI subject.

Concerning the acoustic analysis, the stimuli will be disyllabic nonsense words of the structure ‘pVCV’. The vowel will be /i/ /a/ or /u/ and the consonant will be /p/, /t/ or /s/. In this way we can examine the vowel-to-vowel effects in the context of consonants in two articulation places (bilabial and alveolar) and in two manners (stop and fricative). Stress will also vary its position. This renders 9 different stimuli for each consonant, so there will be 54 different stimuli in total. An example for ‘t’ is the following:

The stimulus will be placed within the carrier phrase ‘Λέγε _____ πάλι’ (Say ____ again’) and each sentence will be repeated 10 times. The 540 sentences will be randomised.

Praat (doing phonetics by computer, copyright ©1992-2004 by Paul Boersma and David Weenink) will be used for the acoustic analysis. F1 and F2 will be measured at the onset, midpoint and end of each vowel so as to monitor the formant movement throughout each disyllable. For the formant onset value, the cursor will be placed at the onset of the second cycle of clear complex periodic activity so as to get a reliable reading from a Gaussian window. The formant offset value will be measured at the end of the cycle with clear complex periodic activity. In addition duration will be measured for each vowel from a waveform display. The left and right cursors will be placed at the onset and offset of complex periodic activity respectively.

The measurements will be analysed statistically for each subject and for each group. The results will attempt to shed light on the coarticulation effects in respect to factors such as: a) the nature of the vowel, b) stress, c) consonant type (place and manner), d) sex (male or female), e) normal vs. impaired hearing, f) degree of hearing-impairment, g) intelligibility score.

For a selected bibliography click here

Some of the acoustic data

You can see the spectrograms of two nonsense words used in the experiment described above: /'papi/ and /pa'si/. The hearing impaired productions are on the left and the normally hearing ones on the right. /'papi/ was produced by a female subject, aged 25, with a hearing loss of 110+ dB and /pa'si/ was produced by a female subject, aged 26, with a hearing loss of 103.3 dB (PTA at 200, 1000 and 2000 Hz). The control subject is a female aged 20. Click on the buttons below the spectrograms to hear the productions.

'papi

  

                           'papi_Hearing Impaired                                                      'papi_Normally Hearing


pa'si

  

                           pa'si_Hearing Impaired                                                      pa'si_Normally Hearing

On the table below you can click and listen to four words and a phrase produced by ten subjects with profound hearing loss (last column, in dB at 500, 1000 and 2000 Hz). This material is part of the intelligibility test described above.

The transcription is in Greek SAMPA symbols.

word 1	['fusces]	balloons
word 2	[fe'gari]	moon
word 3	['dzaci]	fireplace
word 4	[ku'mbja]	buttons
phrase	[min a'gjiksis to ti'Gani jja'ti 'cei]	Don't touch the frying-pan bacause it's too hot.

word 1	word 2	word 3	word 4	phrase	Sex	dB
HI_01_word1	HI_01_word2	HI_01_word3	HI_01_word4	HI_01_phrase	F	101.7
HI_02_word1	HI_02_word2	HI_02_word3	HI_02_word4	HI_02_phrase	F	101.6
HI_03_word1	HI_03_word2	HI_03_word3	HI_03_word4	HI_03_phrase	M	103.3
HI_04_word1	HI_04_word2	HI_04_word3	HI_04_word4	HI_04_phrase	F	110+
HI_05_word1	HI_05_word2	HI_05_word3	HI_05_word4	HI_05_phrase	M	100.7
HI_06_word1	HI_06_word2	HI_06_word3	HI_06_word4	HI_06_phrase	F	103.3
HI_07_word1	HI_07_word2	HI_07_word3	HI_07_word4	HI_07_phrase	M	98.3
HI_08_word1	HI_08_word2	HI_08_word3	HI_08_word4	HI_08_phrase	M	99.3
HI_09_word1	HI_09_word2	HI_09_word3	HI_09_word4	HI_09_phrase	F	96.7
HI_10_word1	HI_10_word2	HI_10_word3	HI_10_word4	HI_10_phrase	M	98.3