Tone Perception And Production As Well As The Vocal Singing Ability Of The Children With Cochlear Implant

Part I The tone perception ability of Mandarin-speaking prelingually-deafened children with cochlear implantObjectiveThis part of study is to explore the tone perception ability of the Mandarin-speaking children who received cochlear implant as well as to find out the potential impact factors underling their tone perception ability, sp as to provide some practical enlightenment for their rehabilitation.MethodsOne hundred and twenty-eight prelingually deafened children who received unilateral cochlear implant participate into this experiment. As control group,154peer children with normal hearing were also recruited.36Chinese characters were selected as test words according to tone balance principle. All of these characters are very simple and in the level of recognizability by children in kindergarten and first several stages of preliminary school. These test words were randomly presented based on MATLAB program in the test procedure. There were a total of four signal-to-noise ratio (SNR) conditions, i.e. quiet,12dB,6dB, and OdB. The subjects can select some or all of these four conditions to administer the tone perception task. Meanwhile, the Pearson correlation analysis was used to evaluate the relationship between their tone perception ability and the demographic factors. Results1. The average tone perception score for NH and CI group were97.78%and77.48%, respectively, the difference was of great statistical significance (P<0.001). For CI group, the tone recognition ability varied dramatically across individuals, ranging from chance level (50%) to100%correct. 2. In the test condition of12,6, and0dB SNR, the average scores of NH and CI group were [94.38%,74.22%],[94.47%,67.43%],[92.31%,59.17%], respectively. The difference between groups were statistically significant (all P<0.001).3. SNR, the tone category (or the tone-pair type), and hearing condition all have a significant effect on tone perception (all P<0.001), among these three factors, hearing condition is extremely outstanding (F=1048.81, far beyond the other two factors). And hearing condition interacted significantly with the other two factors.4. The effect of SNR. For NH group, the score in quiet or6dB SNR was significantly higher than in0dB SNR (both P<0.05), while the scores in all other SNR conditions were not significantly different from each other (all P>0.05). For CI group, the score in each SNR condition was significantly different from in all other conditions (all P<0.05), sorting from highest to lowest was as quiet>12dB>6dB>0dB.5. The effect of tone category. For NH group, tone1or tone4was significantly better recognized than tone2or tone3(P<0.05), while there was no significant difference either between tone1and tone4or between tone2and tone3(both P>0.05). For CI group, except that tone1was significantly better recognized than tone4(P<0.05), no differences were seen statistically between any other two tones (all P>0.05).6. The effect of tone-pair type. For NH group,1-4tone pair>1-2tone pair>2-3tone pair (P<0.05) according to the perception scores, while no statistical differences were seen between any other two tone pairs (all P>0.05). For CI group, the perception score of2-3tone pair was the lowest and was significantly lower than that of1-2,1-3, or1-4tone-pair (all P<0.05). Any other two tone-pairs did not show statistical difference (all P>0.05). 7. The perception confusion pattern. For NH group, tone2and tone3was the most confusing tone pair in all SNR conditions. Tone3perceived as tone2was the most frequently seen error. For CI group, the confusion pattern was influenced by the SNR conditions, the errors diffused among all tone categories. In general, tone3always tends to be perceived as tone2.8. Correlation analysis. For CI group, tone perception score in quiet was negatively correlated with age of implantation (r=-0.392, P<0.001), and tone perception score in12dB SNR was positively correlated with CI use duration (r=0.258, P=0.047). All other correlations were not of statistical significance (all P>0.05).Conclusions1. Prelingually deafened children with CIs have a dramatic deficit in tone perception, their tone perception ability is susceptible to noise environment. Additionally, marked variances do exist across individuals.2. The tone category seems pose a minor effect for children with CIs on thejr perception performance. However,2-3tone pair was the most challenging pair to discriminate for both normal-hearing children and implanted children.3. The younger at implantation and the longer duration of CI use, the better perception performance would be for those implanted children. Part Ⅱ The evaluation of tone production ability of Mandarin-speaking children with cochlear implant based on the artificial neural network modelObjective This part of study is to explore the tone production ability of the Mandarin-speaking children who received cochlear implant, to acquire the knowledge of the acoustic features of their tone production as well as to find out the potential impact factors underling their tone production ability. Also to assess the efficacy and reliability of the artificial neural network used to automatically judge the produced tone categories.MethodsTwo hundred and seventy-eight prelingually deafened children who received unilateral cochlear implant participate into this experiment. As control group,173peer children with normal hearing were also recruited. The same36Chinese characters in the first part of this study were selected as eliciting words to elicit the tone production of each subject. These vocal production samples were then recorded using a sampling rate of44.1KHz and a resolution of16bits. Each syllable was isolated and the fundamental frequency (FO) contour of each syllable was extracted using a MATLAB program and manually corrected afterwards. An artificial neural network was created based on MATLAB, and one half of FO data from the normal hearing group was used to train the neural network. When the training was done, the other half of FO data from the normal hearing group and all the FO data from the cochlear implant group were imported to the neural network and judged. The Pearson analysis was applied to evaluate whether there would be a relationship between the judged results of the cochlear implant group by the neural network and the demographic factors. The relationship between tone perception and production of cochlear implant group was assessed either.Results1. The average tone production accuracy rate for CI group was only58.76%, which is significantly lower than90.40%from NH group (P<0.001). Tremendous variance do exist across CI individuals, their scores ranged from lower than chance level (25%) to100%.2. Tone category and hearing condition showed significant effect on tone production accuracy (both P<0.001). The F value is extremely high for hearing condition (641.75), which was far beyond the F value of tone category (66.41). Meanwhile, these two factors interacted significantly in terms of the influence on tone production (P<0.001).3. The effect of tone category. For NH group, the production accuracy of tone1(or4), tone2, and tone3are significantly different from each other (all P<0.05), with the order of tone1(or4)> tone2> tone3in accordance with accuracy rate from highest to lowest. There was no significant difference between the production accuracy of tone1and tone4(P>0.05). For CI group, tone lor tone4were produced better than tone2or tone3(P<0.01), while no difference was seen between tone1and4or tone2and3(P>0.05).4. The tone production error pattern. For NH group, tone2and tone3are the most confusing tones to be judged as each other when vocally produce. Production of tone1and tone4are less confused or judged as other tones. For CI group, tone1was evenly incorrectly judged as the three other tones, while the three other tones were most likely to be judged as tone1.5. The acoustical properties of tone production. For NH group, the four tones were well classified by the four ellipses locating in upper-right, upper-left, lower-left, and lower-right corner of the coordinate and differentiable from each other, while for CI group, the ellipses overlaid with each other and indistinguishable. The three acoustical indices all showed significantly lower values for CI group than for NH group (all P<0.001).6. Correlation analysis. For CI group, the tone production accuracy was negatively correlated with age of implantation (r=-0.215, P=0.003) and positively correlated with CI use duration (r=0.203, P=0.005). The relationship between tone production and chronological age was of no statistical significance (r=0.015, P=0.837). Meanwhile, tone production accuracy was positively correlated with tone perception score in all SNR conditions (all P<0.01).Conclusions1. Prelingually deafened children with CIs have a dramatic deficit in tone production as well as tone perception. However, the performance varied dramatically across individuals.2. Producing tone2and tone3in more challenging than tone1and tone4for both normal-hearing children and those with CIs.3. The produced four tones are indistinguishable acoustically from each other for children with Cis. Their produced fundamental frequency contour tends to be flat and monotonic, therefore sounds like tone1in most cases. Implanting at an earlier age and persistently using the device are beneficial for the tone production ability of implanted children, also, tone perception ability contributes to their tone production.4. Artificial neural network is a promising technical tool for the automatic recognition of tone. Part Ⅲ Acoustic properties of vocal singing in prelingually-deafened children with cochlear implants or hearing aidsObjective The purpose of the present study was to investigate vocal singing performance of hearing-impaired children with cochlear implants (CI) and hearing aids (HA) as well as to evaluate the relationship between demographic factors of those hearing-impaired children and their singing ability. For children with CIs, also explore the relationship between their singing ability and their tone perception and production capacity.MethodsThirty-seven prelingually-deafened children with CIs and31prelingually-deafened children with HAs, and37normal-hearing (NH) children participated in the study. The fundamental frequencies (FO) of each note in the recorded songs were extracted and the duration of each sung note was measured. Five metrics were used to evaluate the pitch-related and rhythm-based aspects of singing accuracy. Meanwhile, these measured results were analyzed by assessing the correlation with demographic factors and the tone perception and production accuracy score obtained from the previous two parts of this dissertation.ResultsChildren with CIs and HAs showed significantly poorer performance in either the pitch-based assessments or the rhythm-based measure than the NH children. No significant differences were seen between the CI and HA groups in all of these measures except for the mean deviation of the pitch intervals. In this measure, the scores of CIs and HAs were2.89±0.82and2.44±0.76respectively, statistically suggesting children with CIs performed poorer than those with HAs for the accuracy of pitch intervals (P<0.05). For both hearing-impaired groups, length of device use was significantly correlated with singing accuracy. For children with CIs, the pitch interval deviation in pitch production was correlated with their tone perception ability (P<0.05). ConclusionsThere is a marked deficit in vocal singing ability either in pitch or rhythm accuracy in a majority of prelingually-deafened children who have received CIs or fitted with HAs. The pitch interval production of children with CIs was related to their tone perception ability. Although an increased length of device use might facilitate singing performance to some extent, the chance of the hearing-impaired children fitted with either HAs or CIs to reach a high proficiency in singing is quite slim.