Not available / text

(Uncorrected OCR) Abstract of thesis entitled Speaker- Independent Recognition of Putonghua Finals submitted by CHAN, Chit Man for the degree of Doctor of Philosophy at the University of Hong Kong � In December 1987 ABSTRACT A detailed study had been performed to address the problem of speaker-independent recognition of Putonghua (Mandarin) finals. The study included 35 Putonghua finals, 16 of which having trailing nasals. They were spoken by 51 speakers: 38 females, 13 males, in 5 different tones for two times. The sample was spectrally analyzed by a bank of 18 nonoverlapping critical-band filters. Three data reduction techniques: Karhunen-Loeve Transformation (KLT) , Discrete Cosine Transformation (OCT) and Stepwise Discriminant Analysis (SDA) , were comparat i vely studied for their feature representation capability. The results indicated that KLT was superior to both OCT and SDA. Furthermore, the theoretic equivalence of OCT to KLT was found to be valid only with 5 or more feature dimensions used in computation. On the other hand, the results also showed that the Hahalanobis and a proposed modified Mahalanobis distance both gave a better measurement of performance than the other distances tested, which included the City Block, Euclidean, Minkowski, and Chebyshev. .,. In the second Part of the study, the Hidden Markov Modelling (HMM) technique was investigated. Three classification methods: Phonemic Labell ing (PL), Vector Quantization (VQ) and a proposed Hybrid Symbol (HS) generation, were studied for use with HMM. Whilst PL was found to be simple and efficient, its performance was not as good as VQ. However, the time taken by VQ was excessive, especially in training. The results with the HS method showed that it .could successfully merge the speed advantage of PL and the better discriminatory power of VQ. An approximately 80% saving in the quantizer training time could be achieved with only a marginal loss in performance. At the same time, it Abs-l Abstract was also found that allowing skipping of states in a Left-to-Right model (LRM) could lead to a negative effect on overall recognition. As an indication of performance, the recognition rate of the simulated system was 81.3%, 95.0% and 98.0% with the best I, 2, and 3 candidates included, respectively, using a 256-level VQ and a 6-state, no-skip LRM on a sample of 8,400 finals from 48 speakers. The specific rates on non-nasal finals achieved even 96% - 98% using the best candidate alone . .. ," Abs-2 / abstract / toc / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

No description available.

No description available.

Se examino la produccion de vocales tonicas y atonas por parte de tres grupos de hablantes de espanol (dos grupos con diferencias en anos de experiencia y un grupo nativo). Los hablantes participaron en un experimento de produccion cuyos datos fueron sometidos a varios analisis. Se encontraron efectos robustos de los anos de practica con una segunda lengua, por el hecho de que los hablantes avanzados fueron capaces de reproducir vocales que no produjeron los intermedios. Para las vocales tonicas, la /e/ es la vocal mas difi­cil de aprender para los hablantes no-nativos, seguida por la /a/ y la /u/, mientras que la /o/ y la /i/ son mas faciles. Para las vocales atonas, la /e/ y la /o/ son las mas difi­ciles de aprender para los hablantes no-nativos, seguidas por la /a/ y la /u/, y finalmente por la /i/, la cual es la mas facil.

Noise reduction aims to improve the quality of noisy speech by suppressing the background noise in the signal. However, there is always a tradeoff between noise reduction and signal distortion--more noise reduction is always accompanied by more signal distortion. An evaluation of the intelligibility of speech processed by several noise reduction algorithms in [23] showed that most noise reduction algorithms were not successful in improving the intelligibility of noisy speech. In this thesis, we aim to utilize acoustic-phonetic knowledge to enhance the intelligibility of noise-reduced speech. Acoustic-phonetics studies the characteristics of speech and the acoustic cues that are important for speech intelligibility. We considered the following questions: what is the noise reduction algorithm that we should use, what are the acoustic cues that should be targeted, and how to incorporate this information into the design of the noise reduction system. A Bayesian noise reduction method similar to the one proposed by Ephraim and Malah in [16] is employed. We first evaluate the goodness-of-fit of several parametric PDF models to the empirical speech data. For classified speech, we find that the Rayleigh and Gamma. with a fixed shape parameter of 5, model the speech spectral amplitude equally well. The Gamma-MAP and Gamma-MMSE estimators are derived. The subjective and objective performances of these estimators are then compared. We also propose to apply a class-based cue-enhancement, similar to those performed in [21]. The processing directly manipulates the acoustic cues known to be important for speech intelligibility. We assume that the system has the sound class information of the input speech. The scheme aims to enhance the interclass and intraclass distinction of speech sounds. The intelligibility of speech processed by the proposed system is then compared to the intelligibility of speech processed by the Rayleigh-MMSE estimator [16] The intelligibility evaluation shows that the proposed scheme enhances the detection of plosive and fricative sounds. However, it does not help in the intraclass discrimination of plosive sounds, and more tests need to be done to evaluate whether intraclass discrimination of fricatives is improved. The proposed scheme deteriorates the detection of nasal and affricate sounds. / Thesis (Master, Electrical & Computer Engineering) -- Queen's University, 2009-08-24 21:32:48.966

With the emergence of Optimality Theory, where the burden of explanation is placed almost entirely on constraints, we have observed in the phonological literature a de-emphasis on the role of structural relationships that hold within and across segments. In this thesis, counter to the current trend, I argue that the most explanatory approach to phonological processes requires reference to highly-articulated representations. I explore a number of phenomena found in the first language acquisition of Quebec French and argue that these phenomena are best captured in an analysis based on structurally-defined markedness, headedness in constituent structure, and relationships between segmental features and their prosodic licensors. / I demonstrate that headedness in constituent structure must be assigned to both input and output forms. In order to encode the dependency relations between input and output representations, I appeal to faithfulness constraints referring specifically to constituent heads. Output representations are regulated by markedness constraints governing complexity within constituents, as well as by licensing relationships that hold between segmental features and different levels of prosodic representation. / At all stages in the development of syllable structure and complex segments, when more than one option is available for the representation of a target string, children select the unmarked option, consistent with the long-held view that early grammars reflect what is unmarked. When input complex structures are reduced in children's outputs, reduction operates in order to ensure faithfulness to the content of prosodic and segmental heads. Finally, in the discussion of consonant harmony, where the French data are supplemented by examples from English, I propose that consonant harmony results from a licensing relation between segmental features and the head of the foot. The differences in foot structure between French and English enable us to account for the contrasts observed between learners of the two languages.

No description available.

This thesis investigates predictions of an exemplar account of syntax, by testing whether manipulating socially salient phonetic detail can alter the grammaticality judgements given to morpho-syntactic constructions in New Zealand English (NZE). Three experiments are were conducted as part of this thesis. The first tested the social saliency of different phonetic variables in NZE, and found phrase final /t/, which can be realised with or without a release, to be strongest. In the second experiment, phrase final /t/ was tested further, and manipulating the release significantly altered both the age and class ratings given to speakers. The way in which it did this reflected the patterns documented in production. In the third experiment, participants were asked to rate the grammaticality of the same sentences. When the results of the previous experiment were included in the statistical model, an effect of the variant came out as significant. The more participants had rated a speaker as older with the released variant in the previous experiment, the less they rated the sentence as grammatical with the released variant. That is, only the most socially salient realisations were able to alter perceived grammaticality. Overall, the results of this thesis suggest that speaker information and pho- netic detail can affect grammaticality judgements. This supports an exemplar model of syntax. Regardless of the theoretical implications of the findings however, the methodological ones are clear. If speakers and realisations of certain phonetic variables can alter grammaticality judgements, then they must be controlled for in the presentation of stimuli to participants.

This study investigates perceptual effects of discharge rate adaptation in the auditory-nerve fibers. Discrimination tests showed that brief synthetic stimuli with stationary formants and periodic source were better discriminated when they had an abrupt as opposed to a gradual onset (non-adapted vs adapted condition). This effect was not observed for corresponding stimuli with noise source. Discrimination among synthetic /da/ stimuli (abrupt onsets) was worse than among /ad/ stimuli when the respective onset and offset frequencies of the second formant (F2) were varied. Similar results were obtained for /ba/ and /ab/. The low discrimination rate in consonant-vowel stimuli (CV) was explained in terms of sensory smearing of spectral information due to rapid formant transitions. Discrimination improved when the smearing effect was reduced by holding the onset formant pattern over a certain period of time of about 1 6ms. The relatively high discrimination score for the VC stimuli was explained by residual masking; extending the VC offset did not improve discrimination. Discrimination of place of articulation in CV syllables was examined in the light of sensory smearing. Two continua of /bu-du/ and /ba-da/ utterances were used in discrimination and identification experiments. It was observed that the discrimination peak for /Cu/ was displaced from the /b/-/d/ boundary, towards a flat F2 transition, suggesting that optimal place discrimination is related to the stability of the auditory representations generated at onset. This result is discussed in relation to current views of categorical perception. / För att köpa boken skicka en beställning till exp@ling.su.se/ To order the book send an e-mail to exp@ling.su.se