In linguistic research, current practices of recording audio data involves a mixture of analogue and digital technologies giving little forethought to methodological issues and insight into the consequences of using one technology over the other. Analogue recorders are inherently flawed and are not ideal for acoustic analysis. With the introduction of digital recorders into the consumer market, it is difficult to decide which recorder is best suited for linguistic and acoustic analysis. The MiniDisc, introduced by Sony in 1992, is a magneto-optical recorder that offers 'CD-like' quality, in a compact and shock-resistant system, thus making it a potential tool for linguistic field research. However it is not known how the psycho-acoustically based compression system, ATRAC, affects the spectrum of speech sounds. This thesis compares the MiniDisc with a DAT recorder and analyses the differences using spectral and spectrographic analyses. In addition, this thesis tries to establish sound practices for using the MiniDisc in linguistic research.

Animals’ long-distance acoustic signal structure minimizes habitat-specific attenuation and distortion, but it is unknown how environmental acoustics shape the signals of dependent young, or short-range signals generally. I investigated the influence of the nest environment on nestling tree swallows’ begging calls by relating nest reverberation and resonance to nest structure; relating call features to nest reverberation, resonance and structure; and testing whether call structure reduced distortion in the home nest. Reverberation was stronger in wider cavities with intact ceilings. Nestlings used shorter calls in more reverberant cavities, and longer calls with higher middle frequency and lower minimum frequency in wider cavities with smoother walls, but did not adjust call frequency in relation to resonance. Calls originally produced in a given nest did not transmit with less distortion than calls originally produced in other nests. These findings suggest that the nest environment may shape the structure of begging calls.

No description available.

No description available.

One of the most important problems to be faced when building a speech recognizer is the lack of sufficient training data to estimate all of the parameters of the system. One of the most successful techniques addressing this problem is to decrease the number of parameters in the system by the use of classification-tree based acoustic models. In this thesis, several variations on this theme will be presented, the first few ones dealing with trees classifying whole HMMs (tree units), the others with trees classifying states of HMMs (tree states). Results obtained on the Air Travel Information System (ATIS) task show that using classification trees produces systems with up to 90% less models (and proportionally less parameters to estimate) with no loss of performance.

In-line monitoring of die-casting of aluminum (A356, A357, 86S), magnesium (AZ91) and a metal matrix composite (Gra-NiRTM 6S:3G) will be monitored using the reflection coefficient obtained by using an ultrasonic technique, the pulse/echo. For each of the materials enumerated above, the average temperature of the mold through its thickness, the end of filling of the part, the solidification of the part in the cavity of the mold, the gap and/or the detachment of the part, the sound velocity and the attenuation of the material will be measured by this ultrasonic technique during the process. / For the materials the melt temperature will not exceed 600°C because the casting is made at the semi-solid state, between the solidus and liquidus, of the materials. A novel high performance buffer rod with a cooling system is integrated into the die. Therefore, ultrasonic measurements can be carried out with high signal-to-noise ratio at elevated temperatures.

No description available.

No description available.

No description available.

No description available.