This thesis concerns the potential use of auditory virtual reality (A VR) in safety-critical situations. Localization accuracy is essential in many VR situations, such as simulated cockpits, where vision is fully occupied and targets must be signified acoustically. However, the errors reported for localizing 3D sounds varies considerably in the literature and some (e.g. Wightman & Kistler, 1989; Wenzel et aI, 1993) report fairly large errors. This thesis consists of an evaluation of the use of acoustic cues to indicate the location of certain targets. A Knowles Electronic Manikin for Acoustic Research (KEMAR) was used to examine the effects of individualized pinnae on localization accuracy. The results showed that using our own pinnae over foreign pinnae provides little or no benefit. More surprisingly, substantial errors were observed in this study. This initial result drove the fundamental investigation into the large angle errors. The method of eliciting subject responses was investigated. The findings established response method as an important methodological feature in localization experiments from the significant effect it has on the results. Error values can be halved when using a categorical method, compared to an unguided (non-categorical) method, possibly because it constrains the subjects' response options. A further possible constraint on subject responses is the effect of memory in absolute judgement tasks. If the memory of one sound impinges on subsequent sounds then the subject's judgement is constrained and the measurement of error may be contaminated. This effect was studied by introducing variable delays that should affect memory .(0 a different extent. No obvious differences in accuracy were noted. This rules out 'interstimulus interval' as a cause for the variability of reported angle errors. Stimulus types were varied in an effort to maximise acuity. Although broadband sounds are purported to give the smallest errors (e.g. Stevens & Newman, 1936; Sandel et aI, 1955), this investigation offered a unique comparison of long and shortduration broadband and complex sounds. But consistently high angle errors forced the inclusions of non-acoustic cues such as vision and head movements, which decreased the error to between 0° and 7°. III The implicatioos for VR in light of the importance of vision (demonstrated by this work), are that it is 'not advisable to implement an auditory cueing system that may conflict or fail to be guided by vision. Where high levels of accuracy are required, as is paramount in safety-criticalsituations, auditory localizatioo is not sufficient as a sole cue to target location. Scientific conclusion: The acoustic cues alone (independent of context) cannot support accurate auditory localization. Applications conclusion: It is not advisable to implement an auditory cueing system that is not guided by vision.

Thesis (M.S.)--Ohio University, March, 1999. / Title from PDF t.p.

Thesis (M.S.)--West Virginia University, 2006. / Title from document title page. Document formatted into pages; contains xi, 182 p. : ill. (some col.). Includes abstract. Includes bibliographical references (p. 151-156).

<p> Lambda carrageenan when injected subcutaneously causes the formation of a connective tissue granuloma. Initially there is a proliferation of connective tissue elements up to about fourteen days followed by regression so that by six weeks most collagenous tissue has disappeared and been replaced by adipose tissue. Lambda carrageenan has been identified in the granuloma by staining reactions with toluidine blue and other stains for acid polysaccharides. The present study was undertaken to localize the carrageenan by means of fluorescent antibody and autoradiography. For this purpose labelling of carrageenan by photoassimilation of 14CO2 into carrageenan was done. Different parameters affecting the incorporation of 14C into the carrageenan fractions were studied.</p> / Thesis / Master of Science (MSc)

Recurrent Neural Networks (RNNs) are pivotal in deep learning for time series prediction, but they suffer from 'exploding values' and 'gradient decay,' particularly when learning temporally distant interactions. Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) have addressed these issues to an extent, but the precise mitigating mechanisms remain unclear. Moreover, the success of feedforward neural networks in time series tasks using an 'attention mechanism' raises questions about the solutions offered by LSTMs and GRUs. This study explores an alternative explanation for the challenges faced by RNNs in learning long-range correlations in the input data. Could the issue lie in the movement of the representations - how hidden nodes store and process information - across nodes instead of localization? Evidence presented suggests that RNNs can indeed possess "moving representations," with certain training conditions reducing this movement. These findings point towards the necessity of further research on localizing representations.

No description available.