No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

Hindcasts from the Directional Ambient Noise Estimation System (DANES) model were compared to in situ ambient noise measurements to determine the accuracy of the U.S. Navy's ambient noise model. One hundred fifty eight (158) sonobuoy ambient noise measurements were acquired at eight locations in the Northeast Pacific (NEPAC) Ocean during November and December 1980. For each sonobuoy observation a DANES hindcast was made using archived fields from Fleet Numerical Oceanography Center for the simultaneous time and location. The difference between the predictions and measurements was calculated. A maximum mean error of 4.9 dB occurred at 200 Hz which appeared to be due to errors in the DANES Historical Temporal Shipping (HITS) data base. The model was insensitive to synoptic shipping, sound speed profiles and wind field inputs when the HITS data base was utilized.

This thesis describes a number of compositions in which the objective was to investigate whether, and how far, the organisation of light can function as an extension of musical expression in the non-figurative visual realm. I explore the extent to which sound and light are compatible as media, in the sense of both being able to communicate a common set of ideas. The thesis begins by placing the discussion in a historical context, with an overview of the history of analogies between sound and light from Antiquity to the 19th century, as well as the history of Light Art. The second part of the thesis describes synaesthesia as a historically developed aesthetic concept and as a field of research that reveals interesting facts about the neuronal processing of stimulations from the senses. The third part forms the core of the research. It leads from a general historic discussion to more specific problems that emerged in my own work with sound and light. Light is a medium strongly characterised by purity; at first, light therefore seemed an inappropriate medium in which to offer plausible translations of different degrees of sonic noise. However, because of the importance that the inclusion of noise has taken in music since the 20th century, this would have meant a severe handicap in looking for a homological relationship between sound and light in artistic contexts. From a discussion of the broad implications the idea of dirt has in social and cultural contexts, the focus is eventually reduced to the aesthetic problem at hand. By means of a classification of three different sorts of noise, a more differentiated understanding becomes possible of the various functions that noise can have. Corresponding forms of ‘dirty light’ eventually become conceivable and artistically applicable. In the fourth part, six compositions and one audiovisual installation are discussed. Each of these works explores different relationships between the visual and sonic component. When appropriate, the various concepts of ‘dirty light’ that have been derived in the third part are reflected in the form of concrete examples. After discussing each work individually, certain practical problems are addressed that surfaced repeatedly under different performance circumstances. In the fifth part I pose the question of how far events that are conceived to be musical have to be based on sonic events. Common definitions of music that describe sonic events as its exclusive concern are questioned and a number of examples of music are discussed where the sonic outcome is hardly audible or even completely silent. I propose a notion that conceives music as a larger field of activity in which visual manifestations form an integral part. The seven audiovisual works form the practical component of this dissertation. As a result of this research a more differentiated understanding of the nature of the coupling of sound and light has emerged, alongside a comprehension of the at times strongly differing views on the general nature of cross-disciplinary works.

The thesis discusses the channel modelling and relay techniques in powerline communications (PLC) which is considered as a promising technology for the Smart Grid communications, Internet access and home area network (HAN). In this thesis, the statistical PLC channel characteristics are investigated, a new statistical channel modelling method is proposed for the in-door PLC. Then a series of the relay protocols are suggested for the broadband communications over power grid. The statistical channel modelling method is proposed to surmount the limits of the traditional deterministic PLC channel models such as multipath model and transmission line model. To develop the channel model, the properties of the multipath magnitudes, interval between the paths, cable loss and the channel classification are investigated in detail. Then, each property is described by statistical distribution or formula. The simulation results show that the statistical model can describe the PLC channels as accurate as deterministic models without the topology information which is a time-consuming work for collecting. The relay transmission is proposed to help PLC adapting the diverse application scenarios. The protocols covers the main relay aspects which include decode/amplify forwarding, single/ multiple relay nodes, full/half duplex relay working mode. The capacity performance of each protocol is given and compared. A series of the facts which improve the performance of the PLC networks are figured out according to simulation results. The facts include that the decode-and-forward is more suitable for the PLC environment, deviation or transforming station is better location for placing relay node and full duplex relay working mode help exploiting the capacity potential of the PLC networks. Some future works are pointed out based on the work of statistical channel model and relay. In the last part of this thesis, an unit based statistical channel model is initialled for adapting various PLC channel conditions, a more practical relay scenario which contains multiple data terminals is proposed for approaching the realistic transmission scenario. At last, the relay for the narrowband PLC Smart Grid is also mentioned as future research topic.

In most of state-of-the-art speech recognition systems, Gaussian mixture models (GMMs) are used to model the density of the emitting states in the hidden Markov models (HMMs). In a conventional system, the model parameters of each GMM are estimated directly and independently given the alignment. This results a large number of model parameters to be estimated, and consequently, a large amount of training data is required to fit the model. In addition, different sources of acoustic variability that impact the accuracy of a recogniser such as pronunciation variation, accent, speaker factor and environmental noise are only weakly modelled and factorized by adaptation techniques such as maximum likelihood linear regression (MLLR), maximum a posteriori adaptation (MAP) and vocal tract length normalisation (VTLN). In this thesis, we will discuss an alternative acoustic modelling approach — the subspace Gaussian mixture model (SGMM), which is expected to deal with these two issues better. In an SGMM, the model parameters are derived from low-dimensional model and speaker subspaces that can capture phonetic and speaker correlations. Given these subspaces, only a small number of state-dependent parameters are required to derive the corresponding GMMs. Hence, the total number of model parameters can be reduced, which allows acoustic modelling with a limited amount of training data. In addition, the SGMM-based acoustic model factorizes the phonetic and speaker factors and within this framework, other source of acoustic variability may also be explored. In this thesis, we propose a regularised model estimation for SGMMs, which avoids overtraining in case that the training data is sparse. We will also take advantage of the structure of SGMMs to explore cross-lingual acoustic modelling for low-resource speech recognition. Here, the model subspace is estimated from out-domain data and ported to the target language system. In this case, only the state-dependent parameters need to be estimated which relaxes the requirement of the amount of training data. To improve the robustness of SGMMs against environmental noise, we propose to apply the joint uncertainty decoding (JUD) technique that is shown to be efficient and effective. We will report experimental results on the Wall Street Journal (WSJ) database and GlobalPhone corpora to evaluate the regularisation and cross-lingual modelling of SGMMs. Noise compensation using JUD for SGMM acoustic models is evaluated on the Aurora 4 database.