This thesis research is focused on using subband and multi rate adaptive signal processing techniques in order to develop practical speech enhancement algorithms. This thesis comprises of research on three different speech enhancement applications. Firstly, design of a novel method for attenuation of a siren signal in an emergency telephony system (by use of single source siren noise reduction algorithms) is investigated. The proposed method is based on wavelet filter banks and series of adaptive notch filters in order to detect and attenuate the siren noise signal with minimal effect on quality of speech signal. Results of my testing show that this algorithm provides superior results in comparison to prior art solutions. Secondly, effect of time and frequency resolution of a filter bank used in a statistical single source noise reduction algorithm is investigated. Following this study, a novel method for improvement of time domain noise reduction algorithm is presented. The suggested method is based on detection of transient elements of speech signal followed by a time varying signal dependent filter bank. This structure provides a high time resolution at points of transient in a noisy speech signal hence temporal smearing of the processed signal is avoided. Additionally, this algorithm provides high frequency resolution at other times which results in a good performing noise reduction algorithm and benchmarking results against a prior art algorithm and a commercially available noise reduction solution show better performance of proposed algorithm. The time domain nature of algorithm provides a low processing delay algorithm that is suitable for applications with low latency requirement such as hearing aid devices. Thirdly, a low footprint delayless subband adaptive filtering algorithm for applications with low processing delay requirement such as echo cancellation (EC) in telephony networks is proposed. The suggested algorithm saves substantial memory and MIPS and provides significantly faster convergence rate in comparison with prior art algorithms. Finally, challenges and issues for implementation of real-time audio signal processing algorithms on DSP chipsets (especially low power DSPs) are briefly explained and some applications of research conducted in this thesis are presented.

The packaging of implanted devices is crucial if high reliability is to be achieved with an intended operating lifetime of decades. The traditional approach with a welded metal enclosure is too large for many of the new devices that are being developed. Wafer bonding, which allows for simultaneously packaging of all the small devices on a wafer, is becoming a cost effective key technology for packaging of microelectromechanical systems (MEMS) and chip scale devices. A number of bond processes exist that have been utilized for packaging applications. However, these packaging methods are usually application specific. This thesis explores a method, which has the potential to become a standard MEMS/CMOS packaging technology. A gold-silicon bond process was developed to bond cap wafers to device wafers. The process, initially applied on experimental wafers, was designed in such a way that it can be applied, without modification, on CMOS processed wafers. The key steps in this technique are post-processing of device wafer, to define gold seal rings around the device area, fabrication of the cap wafer with corresponding silicon seal rings, and finally bonding these two wafers together according to the developed bond process. Various characterization techniques were employed to quantify and qualify the bonds, including reliability and accelerated aging. This thesis discusses the fabrication process for the developed technique and the challenges associated with all the steps of wafer fabrication and bonding. The result of this study indicated that the developed process can be employed for micropackaging of chip scale devices including MEMS and biomedical implants. An important feature is the ability to incorporate a humidity sensor as part of the active device. Thus, a thin film humidity sensor was also designed, fabricated and characterized as part of this research work. The sensor allows the internal humidity to be measured after the sealed device is implanted thus avoiding the danger of using the device if moisture gets in. Careful attention was paid to the effect of processing steps, in particular etching process, on the sensor performance. A set of experiments was designed to investigate the modifications in physical and chemical characteristics of the moisture sensing film when subjected to wet and dry etching.

This thesis investigates optical fibre transmission system technologies, and their impact on network architectures with the objective of lowering unit cost ($/Gb/s/km) of data transmission in long-haul, and ultra long-haul dense wavelength division multiplexing (DWDM) photonic networks. The importance of this work is driven by the exponential growth in Internet traffic of around 40% p.a., and economic pressures constraining network operators’ ability to invest in their networks. Optical transport networks must therefore be designed to meet future bandwidth demands of end users, with optimum performance, cost and flexibility. Dynamic gain equalisers (DGEs) are a key sub-system of ultra long-haul networks, enabling increased un-regenerated transmission reach and elimination of expensive optical-electrical-optical (OEO) regeneration. A theoretical framework was developed integrating models of wideband power variation, together with narrowband nonlinear propagation simulations using the split-step Fourier method. The optimum spacing of the also costly DGEs was determined for a 3,000km network field deployment. Optimum power pre-emphasis profiles were predicted and compared with simple linear calculations, showing <0.7dB performance penalty using the much faster, simplified method. Optical dispersion management schemes were studied, with optical dispersion compensating fibre placed after every other span resulting in 6% cost reduction and little performance degradation compared to compensation after every span. A techno-economic comparison of optical and electronic dispersion compensation (EDC) strategies showed 25% cost reduction using EDC. Tolerance to fibre nonlinearities is reduced compared to optical compensation; splitting the EDC function equally between transmitter and receiver optimises performance. Economic benefits of a single flexible, multi-reach DWDM system were investigated showing almost 20% cost savings compared to separate long-haul and ultra long-haul systems. Finally, the techno-economic benefits of optical bypass in meshed networks were analysed for increasing levels of optical transparency: from OEO regenerated to multi-degree reconfigurable optical add-drop multiplexers (MD-ROADMs), enabling up to 46% cost saving.

This thesis, submitted for the degree of Doctor of Philosophy at University College London, is a discussion and analysis of combined stepped-frequency and pulse-Doppler target recognition methods which enable a multifunction phased array radar designed for automatic surveillance and multi-target tracking to offer a Non Cooperative Target Recognition (NCTR) capability. The primary challenge is to investigate the feasibility of NCTR via the use of high range resolution profiles. Given stepped frequency waveforms effectively trade time for enhanced bandwidth, and thus resolution, attention is paid to the design of a compromise between resolution and dwell time. A secondary challenge is to investigate the additional benefits to overall target classification when the number of coherent pulses within an NCTR wavefrom is expanded to enable the extraction of spectral features which can help to differentiate particular classes of target. As with increased range resolution, the price for this extra information is a further increase in dwell time. The response to the primary and secondary challenges described above has involved the development of a number of novel techniques, which are summarized below: • Design and execution of a series of experiments to further the understanding of multifunction phased array Radar NCTR techniques • Development of a ‘Hybrid’ stepped frequency technique which enables a significant extension of range profiles without the proportional trade in resolution as experienced with ‘Classical’ techniques • Development of an ‘end to end’ NCTR processing and visualization pipeline • Use of ‘Doppler fraction’ spectral features to enable aircraft target classification via propulsion mechanism. Combination of Doppler fraction and physical length features to enable broad aircraft type classification. • Optimization of NCTR method classification performance as a function of feature and waveform parameters. • Generic waveform design tools to enable delivery of time costly NCTR waveforms within operational constraints. The thesis is largely based upon an analysis of experimental results obtained using the multifunction phased array radar MESAR2, based at BAE Systems on the Isle of Wight. The NCTR mode of MESAR2 consists of the transmission and reception of successive multi-pulse coherent bursts upon each target being tracked. Each burst is stepped in frequency resulting in an overall bandwidth sufficient to provide sub-metre range resolution. A sequence of experiments, (static trials, moving point target trials and full aircraft trials) are described and an analysis of the robustness of target length and Doppler spectra feature measurements from NCTR mode data recordings is presented. A recorded data archive of 1498 NCTR looks upon 17 different trials aircraft using five different varieties of stepped frequency waveform is used to determine classification performance as a function of various signal processing parameters and extent (numbers of pulses) of the data used. From analysis of the trials data, recommendations are made with regards to the design of an NCTR mode for an operational system that uses stepped frequency techniques by design choice.

Peer-to-Peer (P2P) applications have been associated with inefficient operation, interference with other network services and large operational costs for network providers. This thesis presents a framework which can help ISPs address these issues by means of intelligent management of peer behaviour. The proposed approach involves limited control of P2P overlays without interfering with the fundamental characteristics of peer autonomy and decentralised operation. At the core of the management framework lays the Active Virtual Peer (AVP). Essentially intelligent peers operated by the network providers, the AVPs interact with the overlay from within, minimising redundant or inefficient traffic, enhancing overlay stability and facilitating the efficient and balanced use of available peer and network resources. They offer an “insider‟s” view of the overlay and permit the management of P2P functions in a compatible and non-intrusive manner. AVPs can support multiple P2P protocols and coordinate to perform functions collectively. To account for the multi-faceted nature of P2P applications and allow the incorporation of modern techniques and protocols as they appear, the framework is based on a modular architecture. Core modules for overlay control and transit traffic minimisation are presented. Towards the latter, a number of suitable P2P content caching strategies are proposed. Using a purpose-built P2P network simulator and small-scale experiments, it is demonstrated that the introduction of AVPs inside the network can significantly reduce inter-AS traffic, minimise costly multi-hop flows, increase overlay stability and load-balancing and offer improved peer transfer performance.

This thesis addresses the mathematical modelling of end-to-end packet delay for Quality of Service (QoS) provisioning in multi-hop wireless networks. The multi-hop wireless technology increases capacity and coverage in a cost-effective way and it has been standardised in the Fourth-Generation (4G) standards. The effective capacity model approximates end-to-end delay performances, including Complementary Cumulative Density Function (CCDF) of delay, average delay and jitter. This model is first tested using Internet traffic trace from a real gigabit Ethernet gateway. The effective capacity model is developed based on single-hop and continuous-time communication systems but a multi-hop wireless system is better described to be multi-hop and time-slotted. The thesis extends the effective capacity model by taking multi-hop and time-slotted concepts into account, resulting in two new mathematical models: the multi-hop effective capacity model for multi-hop networks and the mixed continuous/discrete-time effective capacity model for time-slotted networks. Two scenarios are considered to validate these two effective capacity-based models based on ideal wireless communications (the physical-layer instantaneous transmission rate is the Shannon channel capacity): 1) packets traverse multiple wireless network devices and 2) packets are transmitted to or received from a wireless network device every Transmission Time Interval (TTI). The results from these two scenarios consistently show that the new mathematical models developed in the thesis characterise end-to-end delay performances accurately. Accurate and efficient estimators for end-to-end packet delay play a key role in QoS provisioning in modern communication systems. The estimators from the new effective capacity-based models are directly tested in two systems, faithfully created using realistic simulation techniques: 1) the IEEE 802.16-2004 networks and 2) wireless tele-ultrasonography medical systems. The results show that the estimation and simulation results are in good agreement in terms of end-to-end delay performances.

This thesis reports on research into the field of Micro-Doppler Signature (μ-DS) based radar Automatic Target Recognition (ATR) with additional contributions to general radar ATR methodology. The μ-DS based part of the research contributes to three distinct areas: time domain classification; frequency domain classification; and multiperspective μ-DS classification that includes the development of a theory for the multistatic μ-DS. The contribution to general radar ATR is the proposal of a methodology to allow better evaluation of potential approaches and to allow comparison between different studies. The proposed methodology is based around a “black box” model of a radar ATR system that, critically, includes a threshold to detect inputs that are previously unknown to the system. From this model a set of five evaluation metrics are defined. The metrics increase the understanding of the classifier’s performance from the common probability of correct classification, that reports how often the classifier correctly identifies an input, to understanding how reliable it is, how capable it is of generalizing from the reference data, and how effective its unknown input detection is. Additionally, the significance of performance prediction is discussed and a preliminary method to estimate how well a classifier should perform is developed. The proposed methodology is then used to evaluate the μ-DS based radar ATR approaches considered. The time domain classification investigation is based around using Dynamic Time Warping (DTW) to identify radar targets based on their μ-DS. DTW is a speech processing technique that classifies data series by comparing them with a pre-classified reference dataset. This is comparable to the common k-Nearest Neighbour (k-NN) algorithm, so k-NN is used as a benchmark against which to evaluate DTW’s performance. The DTW approach is observed to work well. It achieved high probability of correct classification and reliability as well as being able to detect inputs of unknown class. However, the classifier’s ability to generalize from the reference data is less impressive and it performed only slightly better than a random selection from the possible output classes. Difficulties in classifying the μ-DS in the time domain are identified from the k-NN results prompting a change to the frequency domain. Processing the μ-DS in the frequency domain permitted the development of an advanced feature extraction routine to maximize the separation of the target classes and therefore reduce the effort required to classify them. The frequency domain also permitted the use of the performance prediction method developed as part of the radar ATR methodology and the introduction of a na¨ıve Bayesian approach to classification. The results for the DTW and k-NN classifiers in the frequency domain were comparable to the time domain, an unexpected result since it was anticipated that the μ-DS would be easier to classify in the frequency domain. However, the naıve Bayesian classifier produced excellent results that matched with the predicted performance suggesting it could not be bettered. With a successful classifier, that would be suitable for real-world use, developed attention turned to the possibilities offered by the multistatic μ-DS. Multiperspective radar ATR uses data collected from different target aspects simultaneously to improve classification rates. It has been demonstrated successful for some of the alternatives to μ-DS based ATR and it was therefore speculated that it might improve the performance of μ-DS ATR solutions. The multiple perspectives required for the classifier were gathered using a multistatic radar developed at University College London (UCL). The production of a dataset, and its subsequent analysis, resulted in the first reported findings in the novel field of the multistatic μ-DS theory. Unfortunately, the nature of the radar used resulted in limited micro-Doppler being observed in the collected data and this reduced its value for classification testing. An attempt to use DTW to perform multiperspective μ-DS ATR was made but the results were inconclusive. However, consideration of the improvements offered by multiperspective processing in alternative forms of ATR mean it is still expected that μ-DS based ATR would benefit from this processing.

Since their introduction in the 1970s and 1980s superresolution systems for point source parameter estimation have received theoretical attention regarding their potential performance. Two aspects of performance in particular are of interest, the accuracy of the parameter estimation and the resolution achievable. Limitations on performance may be considered to be due to noise affecting the data, or to errors in the system. Superresolution methods divide roughly into two groups – ‘spectral’ methods and maximum likelihood (ML) methods. MUSIC is perhaps the most effective example of a spectral method and has been studied in considerable detail, in both performance measures, but mainly only for the case of a single parameter. In this study the accuracy of MUSIC in the application of two-dimensional direction finding (DF) has been analysed, with and without system errors, using a general array. Theoretical results are confirmed by simulations. An aim has been to produce simpler results for use in estimating the potential performance of practical systems. Little work has been reported on the resolution of ML methods and this is the second main topic of this work, particularly for the two-dimensional DF case using a general array, with a ML method (IMP) similar to the better known Alternating Projection. Some results are obtained for resolution with and without errors for the case of noncoherent signals. For coherent signals (including the standard radar case) the performance is found to depend on the relative phase of the signals, varying from the quadrature case, where the performance is as for the non-coherent case, to the in-phase (or antiphase) case where only one signal peak is seen.

Netted radar employs several spatially distributed transmitters and receivers for information retrieval. This system topology offers many advantages over traditional monostatic and bistatic systems which use a single transmitter and a single receiver. For example, it provides better utilization of reflected energy, more flexible system arrangement and enhanced information retrieval capability. Therefore, the netted radar system is of emerging interests among radar researchers. This work investigates several fundamental aspects that determine netted radar performance. This includes netted radar sensitivity, the netted radar ambiguity function and the netted radar ground plane effect. Mathematical models are developed to provide a mean to examine different aspects of netted radar performance. Software simulations examine netted radar performance over a range of parameter variations. Simulation results show that netted radar can offer better performance over traditional monnostatic and bistatic radar in many cases. Some elementary field trials have been conducted using a prototype netted radar system developed within the UCL radar group to examine aspects of netted radar performance in practice. The field trials are focused on netted radar range and sensitivity which are fundamental. The field trial results show that the theoretical benefits that netted radar can offer are generally realizable in practice.

Intrinsic Josephson Junctions (IJJs) are the subject of much of research due to their potential applications as high frequency oscillators and detectors in the THz range. A number of previous studies use a single junction model to explain their observations, as no fully agreed model for coupled IJJs exists. The influence of one or more IJJs in the voltage state on other junctions in the IJJ array is not yet fully understood. I have studied the dynamical behaviour of Tl_2Ba_2CaCu_2O_8 thin film IJJ arrays, focussing on how one junction in the voltage state influences the other junctions. This I have done by measuring (a) the switching current distributions, and (b) the influence of r.f. irradiation on the DC current-voltage characteristics. I have compared the switching current distributions when switching from the supercurrent branch with those when switching from the first quasiparticle branch. The supercurrent branch was found to be overdamped at the escape frequency whereas the first quasiparticle branch is underdamped, resulting in a larger switching current for the former than the latter. RF irradiation suppresses the mean switching current on the supercurrent branch accompanied by the appearance of a low-voltage flux flow branch. There is however no effect on the switching current of the quasiparticle branches. I explain these results in terms of the dissipative environment in which the junctions are embedded, and, furthermore, that switching of a single IJJ into the voltage state significantly changes the dissipation. IJJs were sub sequently isolated from their environment by FIB-deposited tungsten resistors of resistance between 100 and 500 Ω. The complete resistor-junction-resistor (RJR) structure is multibranched and shows a change in the DC current-voltage characteristics below and above the T^c of the tungsten. However, the presence of high resistance in the RJR structure at 4.2 K needs further investigation.