Andreas Krenmayr

The aim of this study was to develop comprehensive test material for Mandarin tone identification in noise (M-TINT) based on the linguistic characteristics in Chinese. In keeping with both the reliability and efficiency in clinical practice, a primary list consisted of 320 words(80 syllables in four tones each) was designed according to the following principles: shortness of the lists, word familiarity and with a meaning in all four tones. The digital sound file was recorded by a male speaker and a female speaker (both of them are radio broadcaster). The tonal identification material database, which included 288 items (72 loudness-balanced syllables in all 4 Mandarin tones) was established by digital filler and balanced for equal loudness. The complete material was recorded in two CDs in a male version and a female version. The speaker-specific masking noises were generated by filtering Gaussian white noise to the speaker's long-term average speech spectrum (LTASS) and by scaling the masking noises to the same RMS amplitudes, as those of the speech, in order to acquire the effective sound masking. The speaker-specific masking noises could be able or disable during the word presentation. The sound pressure level could be selected concerning on the test setting. The mandarin tonal identification materials were designed by both the tonal acoustic properties and the psychophysics characteristic of adults. It is an useful speech test in clinical work and research, and can potentially be used as the basic list for other tonal language identification test in the future.

To assess the contribution of the temporal fine structure cues on speech recognition, especially tone perception of cochlear implant users whose native language is Mandarin Chinese. Twelve Mandarin speaking users of the MED-EL COMBI 40+ cochlear implants with the standard COMBI 40+ electrode array participated in the study. The subjects were followed up and studied using the longitudinal method for six weeks to compare the difference in speech identification results between fine structure processing (FSP) and continuous interleaved sampler (CIS). The test battery comprised two objective speech tests, including tone identification in quiet as well as a sentence recognition task in quiet and noise. The scores of acute speech test: after replacing FSP strategy, the acute tone test results were similar to that of daily used CIS strategy(male version t = 0.791, P = 0.446;female version t = 0.679, P = 0.511), but the scores of MHINT were decreased significantly (t = -4.365, P = 0.001). The scores of long-lasting speech test were as follows the scores of tone identification test were sharply increased after subjects used the new coding strategy for six weeks continuously(male version t = 3.023, P = 0.018;female version t = 3.883, P = 0.003) and the MHIN results were similar to that of daily used CIS strategy(t = 1.475, P = 0.173). The extraction of temporal fine structure cues supplemented in speech processing strategy of cochlear implant benefits the improvements in tone perception and the understanding of speech in cochlear implant recipients, which can further promotes their quality of life.

The aim of this study was to develop comprehensive test material for Mandarin tone identification in noise for a male and a female talker. Additionally, the sensitivity index d' as a measure for the listeners' performance to identify individual tones was evaluated. The study followed a prospective design. The complete material comprises 72 loudness-balanced syllables in all 4 Mandarin tones. For a selection of 20 syllables, i.e. 80 test words, performance-versus-intensity functions were measured in spectrally matched noise for 16 normal-hearing participants. The average speech reception thresholds in noise were -12.9 dB for the male and -13.6 dB for the female talker recordings. The corresponding slopes were 8.6%/dB and 7.3%/dB. As a performance measure for individual tones, the proportion of correct responses to specific tones was substantially contaminated by response bias. The sensitivity index d', calculated according to detection theory, provided reasonable and unbiased performance versus intensity functions. The results firstly indicate that the material is homogenous enough for use as a speech test in clinical work and research. Secondly, to assess the discrimination performance for individual tones, d' values outperform the simple proportion of correct responses.

Conclusion: Acute comparisons between continuous interleaved sampling (CIS) and a temporal fine structure (TFS) coding strategy in Cantonese-speaking cochlear implant (CI) users did not reveal any significant differences in speech perception. Performance with the unfamiliar TFS coding strategy was on a par with CIS. Benefits of extended fine structure use observed in other studies should be investigated for tonal languages. Objectives: CIS-based stimulation strategies lack an explicit representation of fine structure, which is crucial for tonal language speech perception. The aim of this study was to assess speech recognition with a TFS coding strategy in Cantonese-speaking CI users with no prior fine structure experience. Methods: The fine structure coding strategy encodes TFS on a few apical channels, while the remaining more basal channels carry CIS stimuli. Twelve MED-EL implantees and long-term CIS users participated in a study comparing recognition for Cantonese lexical tones and CHINT sentences between CIS and fine structure stimulation. Results: Mean tone identification scores in 12 subjects were 59.2% with CIS and 59.2% with fine structure stimulation using 4 TFS channels, mean scores of CHINT sentences in 8 subjects were 54.2% with CIS and 55.9% with TFS stimulation. Differences between the two strategies were not significant for any speech test. Two additional versions of TFS strategy and pulse rates were tested in six subjects. No significant differences between strategies were found.