A design of speech recognition system for Chinese names of historical figures around the world is proposed in this thesis. A speech database of approximately forty-six thousand Chinese names is collected and recorded twice for system evaluation. This system applies Mel-frequency cepstrum coefficients, monosyllable HMM¡¦s and speech-text alignment scheme to accomplish initial candidate selection. A Mandarin pitch identification mechanism is then followed to increase the correct rate and obtain the final answer. The experimental results indicate that a 90% correct identification rate can be achieved, under the condition that the first session recording material is used for training and the second one for testing. For the speaker dependent case, the correct name can be recognized within 1.5 seconds, using a PC with an Intel Celeron 2.4 GHz CPU and RedHat Linux 9.0 Operation System.

A four-session text independent, TV-recorded audio-video database for speaker recognition is collected in this thesis. The speaker data is used to verify the applicability of a design methodology based on Mel-frequency cepstrum coefficients and Gaussian mixture model. Both single-session and multi-session problems are discussed in the thesis. Experimental results indicate that 90% correct rate can be achieved for a single-session 3000-speaker corpus while only 67% correct rate can be obtained for a two-session 800-speaker dataset. The performance of a multi-session speaker recognition system is greatly reduced due to the variability incurred in the recording environment, speakers¡¦ recording mood and other unknown factors. How to increase the system performance under multi-session conditions becomes a challenging task in the future. And the establishment of such a multi-session large-scale speaker database does indeed play an indispensable role in this task.

This paper presents text-independent speaker identification results for medium-size speaker population sizes up to 400 speakers for TV speech and TIMIT database . A system based on Gaussian mixture speaker models is used for speaker identification, and experiments are conducted on the TV database and TIMIT database. The TV-Database results show medium-size population performance under TV conditions. These are believed to be the first speaker identification experiments on the complete 400 speaker TV databases and the largest text-independent speaker identification task reported to date. Identification accuracies of 94.5% on the TV databases, respectively and 98.5% on the TIMIT database .

A Mandarin keyword spotting system based on MFCC, discrete-time HMM and Viterbi algorithm with DTW is proposed in this thesis. Joining with a dialogue system, this keyword spotting platform is further refined to a prototype of natural speech patient registration system of Kaohsiung Veterans General Hospital. After the ID number is asked by the computer-dialogue attendant in the registration process, the user can finish all relevant works in one sentence. Functions of searching clinical doctors, making and canceling registration are all built in this system. In a laboratory environment, the correct rate of this speaker-independent patient registration system can reach 97% and all registration process can be completed within 75 seconds.

This thesis investigates the design and implementation strategies for a Taiwanese speech recognition system. It adopts a 4 plus 1¡]five times¡^recording strategy, where the 1st four recordings are used for speech feature training and the last recording for speech recognition simulation. Mel-frequency cepstrum coefficients and hidden Markov model are used as the feature model and the recognition model respectively. Under the Intel Celeron 2.4 GHz personal computer and Red Hat Linux 9.0 operating system environment, a correct phrase recognition rate of 90% can be reached for a 4200 Taiwanese phrase database.

My diploma theses deals with the problem of the speaker recognition. The basic theory of this problem is described in the text as well as model and implementation of the system for speaker recognition. The scope of the system is to recognize up to three speakers. The theory is based on calculation parameters for speaker recognition and processing of voice. Program is made in Matlab as a independent application and it has got Czech and English interface.

Two experiments were conducted to investigate Adaptive Noise Cancelling and Cepstrum echo removal post-processing techniques on acoustic data from a linear microphone array in an anechoic chamber. A point source speaker driven with white noise was used as the primary signal. The first experiment included a background speaker to provide interference noise at three different Signal-to-Noise Ratios to simulate noise propagating down a wind tunnel circuit. The second experiment contained only the primary source and the wedges were removed from the floor to simulate reflections found in a wind tunnel environment. The techniques were applicable to both signal microphone and array analysis. The Adaptive Noise Cancellation proved successful in its task of removing the background noise from the microphone signals at SNRs as low as -20 dB. The recovered signals were then used for array processing. A simulation reflection case was analyzed with the Cepstral technique. Accurate removal of the reflection effects was achieved in recovering both magnitude and phase of the direct signal. Experimental data resulted in Cepstral features that caused errors in phase accuracy. A simple phase correction procedure was proposed for this data, but in general it appears that the Cepstral technique is and would be not well suited for all experimental data. / Master of Science

Knochen und Muskel sind die wichtigsten Gewebe im muskuloskelettalen System welche dem Körper die Bewegungen möglich machen. Beide Gewebetypen sind hochgradig strukturierter Extrazellulärmatrix zugrundegelegt, welche die mechanischen und biologischen Funktionen bestimmen. In dieser Studie wurden die räumliche Verteilung der anisotropen elastischen Eigenschaften und der Gewebemineralisation im humanen kortikalen Femur untersucht mit akustischer Mikroskopie und Synchrotron-µCT. Die homogenisierten elastischen Eigenschaften wurden aus einer Kombination der Porosität und der Gewebeelastizitätsmatrix mit Hilfe eines asymptotischen Homogenisierungsmodells ermittelt. Der Einfluss der Gewebemineralisierung und der Strukturparameter auf die mikroskopischen und mesoskopischen elastischen Koeffizienten wurde unter Berücksichtigung der anatomischen Position des Femurschaftes untersucht. Es wurde ein Modell entwickelt, mit welchem der intramuskuläre Fettgehalt des porcinen musculus longissimus nichtinvasiv mittels quantitativem Ultraschall und dessen spektraler Analyze des Echosignals bestimmt werden kann. Muskelspezifische Parameter wie Dämpfung, spectral slope, midband fit, apparent integrated backscatter und cepstrale Paramter wurden aus den RF-Signalen extrahiert. Die Einflüsse der Muskelkomposition und Strukturparameter auf die spektralen Ultraschallparameter wurden untersucht. Die akustischer Parameter werden durch die Muskelfaserorientierung beeinflusst und weisen höhere Werte parallel zur Faserlängsrichtung als senkrecht zur Faserorientierung auf. Die in dieser Studie gewonnenen detaillierten und lokal bestimmten Knochendaten können möglicherweise als Eingabeparameter für numerische 3D FE-Simulationen. Darüber hinaus kann die Untersuchung von Veränderungen der lokalen Gewebeanisotropie neue Einsichten in Studien über Knochenumbildung geben. Diese auf Gewebeebene bestimmten Daten von Muskelgewebe können in numerischen Simulationen von akustischer Rückstreuung genutzt werden um diagnostische Methoden und Geräte zu verbessern. / Bone and muscle are the most important tissues in the musculoskeletal system that gives the ability to move the body. Both tissues have the highly oriented underlying extracellular matrix structure for performing mechanical and biological functions. In this study, the spatial distribution of anisotropic elastic properties and tissue mineralization within a human femoral cortical bone shaft were investigated using scanning acoustic microscopy and synchrotron radiation µCT. The homogenized meoscopic elastic properties were determined by a combination of porosity and tissue elastic matrix using a asymptotic homogenization model. The impact on tissue mineralization and structural parameters of the microscopic and mesocopic elastic coefficients was analyzed with respect to the anatomical location of the femoral shaft. A model was developed to estimate intramuscular fat of porcine musculus longissimus non-invasively using a quantitative ultrasonic device by spectral analysis of ultrasonic echo signals. Muscle specific acoustic parameters, i.e. attenuation, spectral slope, midband fit, apparent integrated backscatter, and cepstral parameters were extracted from the measured RF echoes. The impact of muscle composition and structural properties on ultrasonic spectral parameters was analyzed. The ultrasound propagating parameters were affected by the muscle fiber orientation. The most dominant direction dependency was found for the attenuation. The detailed locally assessed bone data in this study may serve as a real-life input for numerical 3D FE simulation models. Moreover, the assessment of changes of local tissue anisotropy may provide new insights into the bone remodelling studies. The data provided at tissue level and investigated ultrasound backscattering from muscle tissue, can be used in numerical simulation FE models for acoustical backscattering from muscle for the further improvement of diagnostic methods and equipment.

In der nichtlinearen Datenreihenanalyse hat sich seit etwa 10 Jahren eine Monte-Carlo-Testmethode etabliert, die Theiler-surrogatmethode, mit Hilfe derer entschieden werden kann, ob eine Datenreihe nichtlinearen Ursprungs sei. Diese Methode wird kritisiert, modifiziert und verallgemeinert. Das, was Theiler untersuchen will braucht andere Surrogatmethoden, die hier konstruiert werden. Und das, was Theiler untersucht braucht gar keine Monte-Carlo-Methoden. Mit Hilfe des in der Arbeit eingeführten Begriffs des Phasensignals werden Testmöglichkeiten dargelegt und Beziehungen zwischen den nichtlinearen Eigenschaften der Zeitreihe und deren Phasenspektrum erforscht. Das Phasensignal wird aus dem Phasenspektrum der Zeitreihe hergeleitet und registriert außerordentliche Geschehnisse im Zeitbereich sowie Phasenkopplungen im Frequenzbereich. <br /> <br /> Die gewonnenen Erkenntnisse werden auf das Problem der Polbewegung angewendet. Die Hypothese einer nichtlinearen Beziehung zwischen der atmosphärischen Erregung und der Polbewegung wird untersucht. Eine nichtlineare Behandlung wird nicht für nötig gehalten. / In the nonlinear data analysis there is a popular Monte Carlo Test method due to Theiler (it was established about 10 years ago), the Theiler surrogate method, which tests whether a time series is of a linear origin. This method is being criticized, modified and generalized in this thesis. What Theiler wants to test, needs other surrogate methods, which are constructed here. And what Theiler really tests, does not need Monte Carlo methods. With the help of the concept of the phase signal, that is introduced here, other test options are possible. The phase signal helps also in investigating the relations between the nonlinear characteristics of the time series and their phase spectrum. The phase signal is derived from the phase spectrum of the time series and registers extraordinary events in the time domain as well as phase couplings in the frequency domain. <br /> <br /> These theoretical approches are applied to the problem of polar motion. The hypothesis of a nonlinear relationship between the atmospheric excitation and the pole movement is examined. A nonlinear treatment is not considered necessary.

The industrial world is in constant demand for faster, cheaper and higher quality manufacturing. Robot utilisation and automation has evolved to become a necessary asset to master in order to stay competitive in the global market. With the growing dependency on robots, unexpected downtime and brakedowns can cause devastating loss of revenue. Consequently, this has lead to an increased importance for an accurate condition based way of performing robotic maintenance. As of writing, robots are predominantly maintained through time dependent maintenance. Part replacement is based on statistical models where maintenance is performed without taking the actual robot condition into consideration. As a result an overall level of uncertainty is ensued, where lacking the ability to properly diagnose the robot, also leads to superfluous repairs. Because of the costly impact this has on production, a condition based maintenance approach to robots would yield increased reliability at a lower cost of maintenance. This research focuses on trying to monitor vibrations in a robot, so as to infer about wear and to provide a first step in vibration based Robot Condition Monitoring. This research has been of multidisciplinary nature where robotics, tribology, mechanical component, signal analysis and diagnosis theory have overlapped in several areas throughout the project. The research has provided a vibration baseline and trends of the theoretical bearing defect frequencies for a hypocycloid gearbox installed on an ABB IRB6600 robot. The gearbox was not worn to a level that a severe gearbox degradation was irrefutably detectable and analysable. Accelerometers normally used on wind turbines were used for the project, and are believed to be sufficiently successful in capturing bearing related signals to accredit it for continued use at the preliminary stages of Robot Condition Monitoring development. A worn RV410F hypocycloid gearbox, was dismantled and analysed. Bearings found inside indicate high degrees of moisture corrosion and extensive surface wear. These findings had decisive roles in what future work recommendations where presented. Areas with great potential are condition monitoring through the use of Acoustic Emission and lubrication analysis. Further recommendations include investigating signal analysis techniques such as cepstrum pre-whitening and discrete wavelet transforms.