The mechanical properties of the lips are of crucial importance to the function of a brass instrument. The natural resonance modes must be able to usefully interact with the instrument air column in order to sustain oscillations. Mechanical frequency responses of human and arti cial lips used to play a brass instrument were measured using a high-speed digital video technique in an attempt to classify the true nature of the lipreed. The results revealed the presence of at least two lip modes that exhibited the characteristic outward-inward striking behaviour seen in many in vitro replica lip-reed measurements. The Q-values of the human lip resonances were considerably lower than those seen for the replica lips. Transverse mechanical response measurements were also performed on an in vitro lip-reed to investigate the coupling between the outward and inward striking modes. The two dimensional motion of the lips during full oscillations was investigated. It is shown that a computational four degree-of-freedom model would be required to fully simulate the observed mechanical motion. The uid behaviour downstream from an in vitro vocal fold model was investigated using particle image velocimetry (PIV). A `free jet' con guration with no downstream acoustical coupling was rst investigated. The measurements revealed an unsteady glottal jet ow, consisting of a high velocity jet core, a transitional region of high jet deceleration and a turbulent mixing region. The jet was consistently skewed at angles to the glottal centreline, and appeared to oscillate back and forth across the centreline during the glottal cycle. The behaviour of the jet core was investigated in detail. A temporal asymmetry was observed in the mean velocity across the jet core such that the highest jet velocities were encountered during the closing phase of the vocal folds. The overall jet behaviour also showed a strong turbulent asymmetry between the opening and closing phases. High levels of vorticity and turbulent motion encountered during the closing phase were associated with the deceleration of the jet. Three vocal fold con gurations that included static replicas of the ventricular bands were nally investigated with the aim of characterising the aerodynamic interaction between the ventricular bands and the vocal folds. A marked e ect on the glottal jet was observed for all con gurations. The most physically realistic con guration appeared to stabilise the glottal jet, leading to a reattachment of the jet to the ventricular bands and a subsequent secondary ow separation from the downstream end. The implications of the aerodynamic interaction is discussed, with particular note to its possible relevance to the lip-reed and mouthpiece interaction in brass playing.

The acoustic impedance is a valuable parameter in musical acoustics. Information contained within this frequency-domain parameter can be used to determine the acoustical behaviour of a musical wind instrument: the notes at which the instrument will play; the ease with which a particular note can be played; and the timbre of the instrument. The time-domain version of acoustic impedance - the input impulse response - gives us information on how sound waves propagate within an instrument under playing conditions and how sound is radiated from the open end or from other holes in the bore. Acoustic impedance data can also be used to calculate an accurate profile of the internal structure of the instrument - referred to as a bore reconstruction. This is very useful as the main bore is usually coiled and difficult to measure mechanically. An accurate reconstruction, however, is only possible if the impedance is measured over a large range of frequencies, typically of the order of tens of hertz up to many kilohertz. The bulk of this work follows on from research where the impedance of a short, closed, cylindrically-symmetric tube has been measured experimentally at high frequencies - 1 kHz up to 20 kHz - and compared with theory. The technique used is known as the Two-Microphone-Four-Calibration system, or TMFC system: two microphones are used to monitor the air pressure in the system, and measurement of four closed tubes of different length are required for calibration. The TMFC system has been modified so that impedance data far below 1 kHz (down to 10 Hz) can be attained for a full instrument as well as instrument components - for example trombone mouthpieces or French horn crooks. Suitable algorithms have been developed for processing the impedance data. Obtaining large bandwidth impedance data has allowed the possibility of accurate reconstructions of an instrument's internal profile. The results are compared with plane-wave theoretical models, which are derived in detail, and other well-documented methods of bore and impedance analysis: the acoustic pulse reflectometer (APR), and the brass instrument analysis system (BIAS). An in-depth discussion and analysis of the TMFC results for test objects and instruments of various lengths are presented. Simulations, whereby bore profiles are artificially altered, and a post-processing method utilising transmission matrix theory (TMT), are explored. A variety of orchestral French horn crooks dating from as early as the 18th-Century have been measured using APR and TMFC. A comparison is made between the capabilities of the two systems. Conclusions - of interest from a historical and manufacturing perspective - are drawn. The BIAS has been used to investigate how changes to the bore profile affect the behaviour of trumpets. The impedance of the trumpet is measured using a variety of leadpipes and mouthpieces.

The measurement of speech reception threshold (SRT) is best evaluated in an individual's first language. Currently there is a paucity of linguistically matched materials to measure the SRT of Zulu First Language Speakers (ZFLS) in Kwa Zulu-Natal (KZN). Therefore, the present study focused on developing and assessing a Zulu SRT word list for Zulu First language Speakers (ZFLS) in KZN, according to adapted criteria for SRT in Zulu. In addition, the application of the developed list was evaluated. The study therefore followed a two phase methodological framework. Phase one focused on the development and assessment of the word list. Phase two focused on the application of the word list in a normal hearing population from KZN. Phase one of the study was realized within a descriptive survey design and Phase two was realized using a within- participant quasi experimental design. Phase one included aims one and two of the study. Phase two included aim three of the study. However, each aim had several objectives which were realized consecutively. For aim one of the study, three objectives were achieved i.e. for objective one, 131 common bisyllabic words were identified by two Zulu speaking language interpreters and two tertiary educators. Eighty two percent of these words were described as bisyllabic verbs. The outcome of objective two concluded that 58 bisyllabic verbs were rated as being familiar, phonetically dissimilar and low tone verbs by five linguistic experts, using a three point Likert scale. The agreement among the raters was generally good for each criterion, according to the Kendall's co-efficient of concordance at 95% level of confidence. Two objectives were generated to realize aim two of the study. These included, the measurement of homogeneity of audibility of the 58 words selected in aim one and the acoustic analysis of the words. The findings for the homogeneity of audibility were based on a logistic regression analysis. Thirty normal hearing adult ZFLS (18-25 years) participated in this aim of the study. The mean slope of 50% for 28 words was 5.98%/dB. Therefore, 28 words were measured as being most homogenous. The 28 words were also assessed acoustically. The acoustic analysis indicated that the pitch contours confirm the prosodic pattern of the words selected in terms of Zulu linguistic structure, as the majority of the verbs (89%) indicated a difference in the pitch pattern between the two syllables. Furthermore, trends were noted with regard to the energy contours. The acoustical analysis supports the findings of objective one of aim two. For aim three of the study, twenty six normal hearing adult ZFLS, with functional proficiency in English were assessed. The SRT was measured using the developed Zulu SRT word list. In addition, the SRT was measured using the original CIDW2 list. The Pearson product moment correlation co-efficient was utilized for the measurement of the relationship between the SRT (Zulu) and the Pure Tone Average (PTA). Similarly, the Pearson product moment correlation between the SRT (English) and PTA was obtained. A good relationship between the SRT scores and PTA was reported when both lists were used. However, a stronger correlation between the Zulu SRT and PTA (r=.76) than with the English SRT and PTA (r=0.62) were noted. The results in aim one and aim two of the study highlighted the importance of adapting the criteria for SRT to suit the structure of the language. Aim three confirmed this premise as the implication of a stronger correlation may be related to the familiarity of the stimuli to the Zulu First Language Speaker. The study therefore contributed to both research and clinical implications. Some of the important research implications for the study include: the application of the Zulu SRT word list to a varied clinical population with a hearing disorder or loss, the standardization of the developed Zulu SRT word list on a larger sample, the development of SRT materials in other African languages in South Africa. Important clinical implications of the study include that the findings in the study support the need for speech material to be appropriate to the language of the client and the developed SRT word list in Zulu is applicable to adult ZFLS in KZN. / Thesis (M. Audiology)-University of KwaZulu-Natal, 2006.

Dans cette thèse, nous avons étudié la dynamique d'aimantation selon deux échelles de temps en utilisant la technique pompe-sonde magnéto-optique résolue en temps. A l'échelle de la picoseconde, la précession de l'aimantation est induite par des impulsions acoustiques dans des structures multicouches composées de deux couches ferromagnétique séparées par une couche métallique (Ni/Au/Py) avec différentes épaisseurs. La synchronisation de la précession des couches ferromagnétiques couplées a été observée. La modification de la précession de l'aimantation d'une couche de Ni est due l'interaction d'échange intercouche avec la couche Py. A l'échelle de 50fs, nous avons étudié la dynamique magnéto-optique cohérente, athermale, thermale et la relaxation des charges et des spins dans (Ni, Co et Fe) par impulsions de 11 fs dans un régime de faible perturbation. L'interaction spin-orbite et l'interaction d'échange jouent un rôle significatif dans la désaimantation ultrarapide. / In this thesis, we have investigated the magnetization dynamics at picosecond and femtosecond time scale using time resolved magneto-optical pump probe technique. At picosecond time scale, the magnetization precession is induced by ultrafast acoustic pulses in a three layered structure with two ferromagnetic layers separated by varying thickness of metallic spacer layer (Ni/Au/Py). The magnetization precession dynamics of the Ni layer is modified due to the interlayer exchange interaction with the Py layer and the synchronized precession of the coupied ferromagnetic layers has been observed. At the timescale of 50fs, coherent magneto-optical, non-thermal, thermal and relaxation dynamics of charges and spins in ferromagnetic transition metals (Ni, Co and Fe) is studied by using 11fs optical pulses in a very low perturbation regime. The spin orbit interaction and exchange interaction play a significant role in the demagnetization of the ferromagnetic metals induced by femtosecond pulses.

This thesis presents novel signal processing algorithms for speech and music dereverberation. The proposed algorithms focus on blind single-channel suppression of late reverberation; however binaural and semi-blind methods have also been introduced. Late reverberation is a particularly harmful distortion, since it significantly decreases the perceived quality of the reverberant signals but also degrades the performance of Automatic Speech Recognition (ASR) systems and other speech and music processing algorithms. Hence, the proposed deverberation methods can be either used as standalone enhancing techniques or implemented as preprocessing schemes prior to ASR or other applied systems. The main dereverberation method proposed here is a blind dereverberation technique based on perceptual reverberation modeling has been developed. This technique employs a computational auditory masking model and locates the signal regions where late reverberation is audible, i.e. where it is unmasked from the clean signal components. Following a selective signal processing approach, only such signal regions are further processed through sub-band gain filtering. The above technique has been evaluated for both speech and music signals and for a wide range of reverberation conditions. In all cases it was found to minimize the processing artifacts and to produce perceptually superior clean signal estimations than any other tested technique. Moreover, extensive ASR tests have shown that it significantly improves the recognition performance, especially in highly reverberant environments. / Η διατριβή αποτελείται από εννιά κεφάλαια, δύο παραρτήματα καθώς και την σχετική βιβλιογραφία. Είναι γραμμένη στα αγγλικά ενώ περιλαμβάνει και ελληνική περίληψη. Στην παρούσα διατριβή, αναπτύσσονται μεθόδοι ψηφιακής επεξεργασίας σήματος για την αφαίρεση αντήχησης από σήματα ομιλίας και μουσικής. Οι προτεινόμενοι αλγόριθμοι καλύπτουν ένα μεγάλο εύρος εφαρμογών αρχικά εστιάζοντας στην τυφλή (“blind”) αφαίρεση για μονοκαναλικά σήματα. Στοχεύοντας σε πιο ειδικά σενάρια χρήσης προτείνονται επίσης αμφιωτικοί αλγόριθμοι αλλά και τεχνικές που προϋποθέτουν την πραγματοποίηση κάποιας ακουστικής μέτρησης. Οι αλγόριθμοι επικεντρώνουν στην αφαίρεση της καθυστερημένης αντήχησης που είναι ιδιαίτερα επιβλαβής για την ποιότητα σημάτων ομιλίας και μουσικής και μειώνει την καταληπτότητα της ομιλίας. Επίσης, επειδή αλλοιώνει σημαντικά τα στατιστικά των σημάτων, μειώνει σημαντικά την απόδοση συστημάτων αυτόματης αναγνώρισης ομιλίας καθώς και άλλων αλγορίθμων ψηφιακής επεξεργασίας ομιλίας και μουσικής. Έτσι οι προτεινόμενοι αλγόριθμοι μπορούν είτε να χρησιμοποιηθούν σαν αυτόνομες τεχνικές βελτίωσης της ποιότητας των ακουστικών σημάτων είτε να ενσωματωθούν σαν στάδια προ-επεξεργασίας σε άλλες εφαρμογές. Η κύρια μέθοδος αφαίρεσης αντήχησης που προτείνεται στην διατριβή, είναι βασισμένη στην αντιληπτική μοντελοποίηση και χρησιμοποιεί ένα σύγχρονο ψυχοακουστικό μοντέλο. Με βάση αυτό το μοντέλο γίνεται μία εκτίμηση των σημείων του σήματος που η αντήχηση είναι ακουστή δηλαδή που δεν επικαλύπτεται από το ισχυρότερο σε ένταση καθαρό από αντήχηση σήμα. Η συγκεκριμένη εκτίμηση οδηγεί σε μία επιλεκτική επεξεργασία σήματος όπου η αφαίρεση πραγματοποιείται σε αυτά και μόνο τα σημεία, μέσω πρωτότυπων υβριδικών συναρτήσεων κέρδους που βασίζονται σε δείκτες αντικειμενικής και υποκειμενικής αλλοίωσης. Εκτεταμένα αντικειμενικά και υποκειμενικά πειράματα δείχνουν ότι η προτεινόμενη τεχνική δίνει βέλτιστες ποιοτικά ανηχωικές εκτιμήσεις ανεξάρτητα από το μέγεθος του χώρου.

[pt] O conhecido algoritmo adaptativo denominado least-mean square (LMS) é uma abordagem simples e eficiente para problemas de cancelamento ativo de ruído (ANC). No entanto, na presença de sinais não Gaussianos ou sistemas não lineares, o clássico LMS comumente não alcança um desempenho satisfatório. Por essa razão, um amplo número de técnicas de processamento de sinais adaptativo robustas tem sido investigadas nas últimas décadas. Essa dissertação propõe uma abordagem de filtragem adaptativa robusta para cancelamento ativo de ruído. Em particular, o modelo utiliza a clássica estrutura filtered-X junto ao método desenvolvido neste trabalho, baseado na derivação de uma função tangente hiperbólica exponencial kernel generalizado M estimador (HEKM), o qual alcançou um desempenho ótimo em termos da Redução de Ruído Média (ANR). Os resultados demonstraram o custo-benefício do algoritmo proposto para supressão de diferentes tipos de sinais espúrios na entrada do sistema. / [en] The well-known adaptive algorithm called least-mean square (LMS) is a simple and efficient approach to active noise cancellation application problems. However, in the presence of non-Gaussian noises or non-linear environments, the standard LMS commonly cannot reach satisfactory performance. Therefore, a wide range of robust adaptive processing techniques have been investigated in the last few decades. This thesis proposes a robust adaptive filtering approach for noise cancellation. In particular, the model uses the classical filtered-X framework with the developed method in this research, it is based on hyperbolic tangent exponential generalized Kernel M-estimator function (HEKM), which achieves optimal performance in terms of Average Noise Reduction (ANR). The results demonstrate the cost-effectiveness of the proposed algorithm in suppressing spurious noises in different input systems.