Computational Modelling of Tone Perception Based on Direct Processing of f0 Contours.

It has been widely assumed that in speech perception it is imperative to first detect a set of distinctive properties or features and then use them to recognize phonetic units like consonants, vowels, and tones. Those features can be auditory cues or articulatory gestures, or a combination of both. There have been no clear demonstrations of how exactly such a two-phase process would work in the perception of continuous speech, however. Here we used computational modelling to explore whether it is possible to recognize phonetic categories from syllable-sized continuous acoustic signals of connected speech without intermediate featural representations. We used Support Vector Machine (SVM) and Self-organizing Map (SOM) to simulate tone perception in Mandarin, by either directly processing f0 trajectories, or extracting various tonal features. The results show that direct tone recognition not only yields better performance than any of the feature extraction schemes, but also requires less computational power. These results suggest that prior extraction of features is unlikely the operational mechanism of speech perception.

Bimodal benefits in Mandarin-speaking cochlear implant users for music perception and tone recognition.

BACKGROUND: Cochlear implant (CI) users have difficulty appreciating music and perceiving lexical tones in Mandarin Chinese. Wearing a hearing aid (HA) in the contralateral ear for bimodal hearing may provide additional benefits. OBJECTIVES: To measure the bimodal benefits of music perception and tone recognition and to investigate the relationship between the two in Mandarin-speaking bimodal CI subjects. MATERIALS AND METHODS: Sixteen Mandarin-speaking bimodal CI subjects (aged between 16 and 49 years) participated in the study. Music perception (pitch discrimination, melody discrimination and instrument identification) and lexical tone recognition were tested with electric stimulation (CI alone) or bimodal stimulation (CI + HA). RESULTS: Subjects showed a significant bimodal benefit in tone recognition in quiet and noise, and in all music perception tests. The bimodal benefit for tone recognition in noise was significantly correlated with that of pitch discrimination thresholds and instrument identification scores. CONCLUSION: Mandarin-speaking bimodal CI users achieved better music perception and tone recognition ability with CI + HA than with CI alone. The bimodal benefit of tone recognition was significantly correlated with that of music perception.

Sound localization, speech and tone recognition for stimuli presented from the rear in bilateral cochlear implant users.

OBJECTIVE: To assess any differences in spatial listening ability of cochlear implant recipients when using both or only one of two bilateral cochlear implants (BCIs) for stimuli originating from behind the subject. DESIGN: Twelve loudspeakers were placed in the rear horizontal plane of the subjects to test the sound localisation performance of BCI users and normal-hearing listeners (NHLs) with or without interfering noise. Stimuli were presented via two rear loudspeakers simultaneously during the speech recognition test. In the tone recognition test, another anechoic chamber was used with stimuli presenting from a loudspeaker behind the participants. STUDY SAMPLE: Twenty-seven NHLs and eleven BCI users. RESULTS: Average root-mean-square (RMS) error for the bilateral condition was significantly lower than that for the right and left cochlear implant (CI) conditions with or without interfering noises (p < 0.05). Average speech or tone recognition scores for the bilateral condition and the right and left CI conditions were not statistically significant (p > 0.05). CONCLUSION: Sound localisation with BCIs was significantly more accurate than with either implant alone. Speech and tone recognition scores were not better with two compared to those of one activated implant. Given the small number of subjects, the results should be considered as preliminary.

Mandarin lexical tone recognition in bimodal cochlear implant users.

Objective: To assess the recognition of lexical tones in Mandarin-speaking bimodal cochlear implant (CI) subjects.Design: Lexical tone recognition in quiet and noise (SNR= +5 dB) was measured with electric stimulation (CI alone) or bimodal stimulation (CI + hearing aid (HA)). The recognition and confusion rates of the four tones (T1, T2, T3 and T4) were analysed. Spearman correlation analysis was performed to examine the relationship between hearing levels in the contralateral ear and bimodal benefits.Study sample: Twenty native Mandarin-speaking bimodal CI users, with ages ranging from 16-49 years.Results: Relative to the CI alone, mean tone recognition with the CI + HA improved significantly from 84.1-92.1% correct in quiet (+8 points) and from 57.9-73.1% correct in noise (+15.2 points). Tone confusions between T2 and T3 were the most prominent in all test conditions, and T4 tended to be labelled as T3 in noise. There was no significant correlation between the bimodal benefits for tone recognition and the unaided or HA-aided pure-tone thresholds at 0.25 kHz.Conclusion: Listeners with CI + HA exhibited significantly better tone recognition than with CI alone. The bimodal advantage for tone recognition was greater in noise than in quiet, perhaps due to a ceiling effect in quiet.

The Effect of Temporal Information in Cochlear Implant Analog Sounds on the Tone Recognition of Normal Hearing Young People / 听力学及言语疾病杂志

Objective By comparing the tone recognition rates for different cochlear implant (CI) analog sounds ,the effects of the temporal coding strategy on tone recognition were investigated .Methods The professional announcer read 6 vowel (/a/,/o/,/e/,/i/,/u/,/ü/) of 4 different tones at a normal speed .After adjusting the am-plitude envelope and increasing the fine structure ,the audios with different sampling accuracy below 500 Hz(125 ,250? ?1500 pps) were obtained ,including 288 audio amplitude envelope adjusted and 288 not adjusted .Thirty young participants of normal hearing had tone recognition tests .The results and the tone recognition rates under two dif-ferent temporal properties were compared .Results When the temporal fine structures were combined in considera-tion ,the amplitude envelope as adjusted to match the fundamental frequency (F0) had a significant effect on the tone recognition .The tone recognition rate after amplitude envelope adjusted (80 .22％ ± 16 .32％ ) was higher than before (74 .83％ ± 20 .24％ ) [F(1 ,9)=16 .91 ,P=0 .002] .When the amplitude envelopes were combined in consider-ation ,changing the fine structure of the frequency below 500 Hz had a significant effect on the tone recognition [F (11 ,99)=38 .86 ,P<0 .001] .When the sampling precision was <375 pps ,the tone recognition had improved re-markably with the improvement of sampling precision (P<0 .004) .When the sampling precision was ≥375 pps , the effect on the tone recognition was not significant (P>0 .004) .The interaction between changing amplitude en-velopes and increasing the fine structure had a significant effect on the tone recognition [ F(11 ,99 )= 3 .78 , P<0 .001] .When the fine structure ≤375 pps ,adjustment on the amplitude envelope to increase the information of F0 had a significant impact on the tone recognition (P<0 .05) .When the fine structure was >375 pps ,the difference of the tone recognition before and after the adjustment tended to 0 (P>0 .05) .Conclusion In the CI coding strate-gy ,adjusting the amplitude envelope to get more F0 information improves tone recognition .To a certain degree ,the tone recognition improves when the sampling precision of frequency below 500 Hz is improved .Adjusting the ampli-tude envelope to F0 and increasing the temporal fine structure below 500 Hz can be combined in the same speech coding strategy to optimize the tone recognition .

Perceptual separation of sensorineural hearing loss and auditory neuropathy spectrum disorder.

OBJECTIVES/HYPOTHESIS: The present study aimed to examine whether the response patterns to the chimeric lexical tone tokens, combined with their pure tone audiometry (PTA) results, could separate listeners with sensorineural hearing loss (SNHL) from listeners with auditory neuropathy spectrum disorder (ANSD). STUDY DESIGN: Case-control study. METHODS: Forty-three SNHL subjects and 46 ANSD subjects participated in a Mandarin lexical tone perception test using original and chimeric tone tokens. Ten sets of monosyllables, with four tone patterns for each, were processed through a 16-channel chimeric synthesizer in which a temporal envelope (E) from a monosyllabic word of one tone was paired with a fine structure (FS) from the same monosyllable of other tones. RESULTS: Significantly negative correlations were present between tone perception scores and PTA0.5-4 kHz for both SNHL (P < 0.001) and ANSD (P < 0.001) subjects. Overall, 72.4%, 66.4%, and 46.3% of the tone responses were consistent with FS for the SNHL subjects with mild, moderate, and severe degree of hearing loss, respectively; and 28.4%, 23.1%, and 22.7% were consistent with FS for the ANSD subjects, with the equivalent degree of hearing loss. Similarly, 17.6%, 24.2%, and 37.7% of the tone responses were consistent with E for the SNHL subjects with mild, moderate, and severe degree of hearing loss, respectively; and 45.5%, 44.3%, and 36.5% were consistent with E for the ANSD subjects. CONCLUSION: Subjects with SNHL and ANSD may be separated by representing their FS- and E-consistent tone responses as a function of their pure-tone hearing thresholds. LEVEL OF EVIDENCE: 3b. Laryngoscope, 126:1420-1425, 2016.