PERCEPTION OF VOWELS AND PROSODY BY CI RECIPIENTS IN NOISE

Get Complete Project Material File(s) Now! »

Speech materials

To compare the perception of prosody at sentence level with the recognition of words in a sentence, the recorded speech materials described in section 3.2 of this chapter were combined with SWN generated in a commercial software package for mathematics. The noise had a spectral envelope matching the average power spectral density of the entire set of sentences recorded from each speaker. This was achieved by determining the average spectrum for each speaker across all utterances, and using the envelope of this spectrum as a filter to shape the spectrum of white noise
accordingly. The resulting noise was added to the recorded speech at SNRs of -2, -5 and -8 dB. The sentences that portrayed conditional versus unconditional prosody were distributed across the three test conditions (SNR-2, SNR-5 and SNR-8) so that each listener heard only one version (either conditional or unconditional) of each sentence at each SNR. To measure word recognition, a different group of phonemically matched sentence lists was used at each SNR to avoid familiarisation with the content (see Appendix B for test lists used at each SNR). Word recognition
was tested using three lists of ten sentences each at each SNR.

Participants

Ten listeners participated in this experiment. All participants were young adults (ages 19-25 years), students at a tertiary education institution, native speakers of Afrikaans (the test language), and had normal hearing (pure tone thresholds ≤ 20 dB HL at 250, 500, 1000, 2000, 4000, and 8000 Hz). Informed consent was obtained from each listener prior to testing, and listeners were rewarded at the standard hourly fee specified by the research group.

Test procedure

Participants were seated in a sound-proof booth with the examiner (a qualified audiologist) for the duration of each experiment. Test materials were presented via the external sound card of a personal computer, through an M-Audio EX66 Reference Monitor. Listeners were seated approximately one metre from the loudspeaker, facing it squarely. Materials were presented at 65 dB SPL as measured at the ear level of the participant. This intensity level was selected as it was considered to be a comfortable listening level by the NH participants. The presentation of the test items was controlled by the administrator, using Praat software as an interface.
The first five participants started with the recognition of the phonemically matched sentences, with the female speaker. Testing commenced with 20 practice sentences, followed by three lists of ten sentences each. This was followed by the phonemically matched sentences as read by the male speaker (again 20 practice sentences followed by 30 test sentences). The sentences were presented one by one, and listeners were required to repeat whatever part of the sentence they were able to hear. The test administrator compared the listener’s response to a written version of the sentence that was printed on a test form, and indicated on the test form the number of words in the sentence that were repeated correctly. After all the sentences for both speakers had been presented, the female speaker’s version of the conditional/unconditional sentences were played, starting with three practice items, followed by the male speaker’s version.

READ GEOGRAPHY OF SYRIO-PALESTINE

Results

Prosody recognition performance was calculated as the percentage of sentences for which the prosodic version (conditional/unconditional) was identified correctly. Word recognition scores were calculated as the percentage of words repeated correctly from the three phonemically matched lists. Because the prosody task was a closed set (2AFC) task while word recognition was an open set task, prosody recognition scores were corrected for guessing using Boothroyd’s equation (1988) as specified in Equation (2.1) of Chapter 2. Data from one of the listeners (listener number 2, a female listener) were excluded from the analyses, as this listener was a clear outlier on all the tests, performing considerably worse than all other listeners.
This listener adhered to the selection criteria for the study, but may have misunderstood some of the instructions of the listening tasks. Results are depicted in Figures 3.5 and 3.6, the former showing results from the two speakers separately, and the latter showing results averaged across the two speakers.

CHAPTER 1 INTRODUCTION .
1.1 PROBLEM STATEMENT
1.1.1 Context of the problem .
1.1.2 Research gap
1.2 HYPOTHESIS
1.3 RESEARCH QUESTIONS
1.4 APPROACH AND OBJECTIVES
1.5 CONTRIBUTION
1.6 OVERVIEW OF THE STUDY
CHAPTER 2 LITERATURE STUDY .
2.1 CHAPTER OBJECTIVES
2.2 SPEECH PERCEPTION IN NOISE: CI RECIPIENTS AND NH LISTENERS .
2.3 PROSODY: DEFINITIONS AND FUNCTIONS
2.4 PROSODY IN NOISE
2.5 PROSODY PERCEPTION IN CI RECIPIENTS
2.6 BACKGROUND ON AFRIKAANS .
2.7CONCLUSION
CHAPTER 3 SENTENCE-LEVEL WORD RECOGNITION VERSUS PROSODY RECOGNITION IN NH LISTENERS IN NOISE
3.1 CHAPTER OBJECTIVES
3.2 BACKGROUND
3.3 DEVELOPMENT AND ACOUSTIC ANALYSES OF TEST MATERIALS
3.3.1. Recording and editing of speech material: methods and results
3.3.2 Validation of prosody recognition materials: methods and results .
3.3.3 Acoustic analyses of prosody materials: methods and results
3.3.4 Discussion
3.4 LISTENING EXPERIMENT: PROSODY AND WORD RECOGNITION IN NOISE
3.4.1 Method
3.4.2 Results
3.4.3 Discussion
3.5 CONCLUSIONS
CHAPTER 4 PERCEPTION OF VOWELS AND PROSODY BY CI RECIPIENTS IN NOISE
4.1 CHAPTER OBJECTIVES
4.2 DEVELOPMENT AND ACOUSTIC ANALYSES OF TEST MATERIALS
4.2.1 Background
4.2.2 Recording and validation of speech material: methods and results
4.2.3 Acoustic analyses: methods and results
4.3 LISTENING EXPERIMENTS
4.3.1. Background .
4.3.2 Methods
4.3.3 Results
4.3.4 Discussion
4.4 CONCLUSIONS
CHAPTER 5 PERCEPTION OF PHONEMES AND EMOTIONAL PROSODY BY CI
RECIPIENTS IN NOISE
5.1 CHAPTER OBJECTIVES
5.2 BACKGROUND
5.2.1 Addressing limitations of previous experiments .
5.2.2 Types of speech material
5.3 DEVELOPMENT, VALIDATION AND ACOUSTIC ANALYSES OF SPEECH
MATERIALS: METHODS AND RESULTS
5.4 PILOT LISTENING EXPERIMENT
5.5 MAIN LISTENING EXPERIMENT
5.5.1 Methods
5.5.2 Results
5.6 DISCUSSION
5.6.1 Acoustic analyses
5.6.2 Listening experiment
5.7 CONCLUSIONS
CHAPTER 6 GENERAL DISCUSSION AND CONCLUSION
REFERENCES