Dissertations / Theses on the topic 'Waveform Classification'

A cognitive radar (CR) system is one that observes and learns from the environment, then uses a dynamic closed-loop feedback mechanism to adapt the illumination waveform so as to provide system performance improvements over traditional radar systems. A CR system that performs multiple hypothesis target classification and exploits the spectral sparsity of correlated narrowband target responses to achieve significant performance improvements over traditional radars that use wideband illumination pulses was recently developed. This CR system, which was designed for Gaussian target responses, is extended to non-Gaussian targets. In this thesis, the CR system is generalized to deal effectively with arbitrary non-Gaussian distributed target responses via two key contributions (1) an important statistical expected value operation that is usually evaluated in closed form is evaluated numerically using an ensemble averaging operation, and (2) a powerful new statistical sampling algorithm and a kernel density estimator are applied to draw complex target samples from target distributions specified by both a desired power spectral density and an arbitrary desired probability density function. Simulations using non-Gaussian targets demonstrate very effective algorithm performance. As expected, this performance gain is realized at the expense of increased computational complexity.

Approved for public release; distribution is unlimited
We set out to design an extended target classification scheme while determining the target’s range-and-Doppler location with the use of adaptive waveform for a closed-loop cognitive radar platform. To that end, this work is divided into three objectives: 1) in support of determining range-Doppler locations, we investigate the ambiguity function of the matched waveform called eigenwaveform, 2) in support of target classification, we look at an adaptive waveform technique called probability-weighted eigenwaveform (PWE) and introduce two new waveforms, and 3) we design an integrated range-Doppler map and extended target classification scheme. In this work, we show that the fundamental properties of ambiguity function for extended targets are different when compared to classical waveforms for point targets. We improve on the adaptive waveform called maximum a posteriori PWE and introduce two new waveforms called match-filtered PWE and two-stage PWE. We propose an integrated range-Doppler map and identification scheme for multiple moving extended targets. Performance comparisons in terms of joint probability of identification and determining targets’ range-Doppler locations with traditional wideband waveform and the three PWE-based waveforms are shown. It is shown that the three PWE-based waveforms perform better than the classical wideband waveform.

To enable a new level of connectivity among machines as well as between people and machines, future wireless applications will demand higher requirements on data rates, response time, and reliability from the communication system. This will lead to a different system design, comprising a wide range of deployment scenarios. One important aspect is the evolution of physical layer (PHY), specifically the waveform modulation. The novel generalized frequency division multiplexing (GFDM) technique is a prominent proposal for a flexible block filtered multicarrier modulation. This thesis introduces an advanced GFDM concept that enables the emulation of other prominent waveform candidates in scenarios where they perform best. Hence, a unique modulation framework is presented that is capable of addressing a wide range of scenarios and to upgrade the PHY for 5G networks. In particular, for a subset of system parameters of the modulation framework, the problem of symbol time offset (STO) and carrier frequency offset (CFO) estimation is investigated and synchronization approaches, which can operate in burst and continuous transmissions, are designed. The first part of this work presents the modulation principles of prominent 5G candidate waveforms and then focuses on the GFDM basic and advanced attributes. The GFDM concept is extended towards the use of OQAM, introducing the novel frequency-shift OQAM-GFDM, and a new low complexity model based on signal processing carried out in the time domain. A new prototype filter proposal highlights the benefits obtained in terms of a reduced out-of-band (OOB) radiation and more attractive hardware implementation cost. With proper parameterization of the advanced GFDM, the achieved gains are applicable to other filtered OFDM waveforms. In the second part, a search approach for estimating STO and CFO in GFDM is evaluated. A self-interference metric is proposed to quantify the effective SNR penalty caused by the residual time and frequency misalignment or intrinsic inter-symbol interference (ISI) and inter-carrier interference (ICI) for arbitrary pulse shape design in GFDM. In particular, the ICI can be used as a non-data aided approach for frequency estimation. Then, GFDM training sequences, defined either as an isolated preamble or embedded as a midamble or pseudo-circular pre/post-amble, are designed. Simulations show better OOB emission and good estimation results, either comparable or superior, to state-of-the-art OFDM system in wireless channels.

There is an increasing demand for urban vegetation mapping, and airborne laser scanning (ALS) has the unique ability to provide geo-referenced three-dimensional data useful for mapping of surface features. This thesis examines the ability of full-waveform and discrete return ALS point data to distinguish urban surface features, and represent the three-dimensional attributes of vegetation at different scales in a vector-based GIS environment. Two full-waveform datasets, at a wavelength of 1550 nm, and a discrete return dataset, at 1064 nm, are used. Points extracted from the first full-waveform dataset are classified with k-means clustering and decision tree into vegetation, buildings and roads, based on the attributes of individual points and the relationships between neighbouring points. A decision tree is shown to perform significantly better (74.62%) than k-means clustering (51.59%) based on the overall accuracies. Grass and paved areas could be distinguished better using intensity from discrete return data than amplitude from full-waveform data, both values proportional to the strength of the return signal. The differences in the signatures of surfaces could be related to the wavelengths of the lasers, and need to be explored further. Calibration of intensity is currently possible only with full-waveform data. When the decision tree is applied on the second full-waveform dataset, the backscatter coefficient proves to be a more useful attribute than amplitude, pointing to the need for calibration if a classification method using intensity is to be applied on datasets with different scanning geometries. A vector-based approach for delineating tree crowns is developed and implemented at three scales. The first scale provides a good estimation of the tree crown area and structure, suitable for estimating biomass and canopy gaps. The third scale identifies the number of trees and their locations and can be used for modelling individual trees.

The Cedar Mountain and Dakota Formations are significant gas producers in the southern Uinta Basin of Utah. To date, however, predicting the stratigraphic distribution and lateral extent of potential gas-bearing channel sandstone reservoirs in these fluvial units has proven difficult due to their complex architecture, and the limited spacing of wells in the region. A new strategy to correlate the Cedar Mountain and Dakota Formations has been developed using core, well-log, and 3D seismic data. The detailed stratigraphy and sedimentology of the interval were interpreted using descriptions of a near continuous core of the Dakota Formation from the study area. The gamma-ray and density-porosity log signatures of interpreted mud-dominated overbank, coal-bearing overbank, and channel sandstone intervals from the cored well were used to identify the same lithologies in nearby wells and correlate similar stratal packages across the study area. Data from three 3D seismic surveys covering approximately 140 mi2 (225 km2) of the study area were utilized to generate spectral decomposition, waveform classification, and percent less-than-threshold attributes of the Dakota-Cedar Mountain interval. These individual attributes were combined to create a composite attribute that was merged with interpreted lithological data from the well-log correlations. The overall process resulted in a high-resolution correlation of the Dakota-Cedar Mountain interval that permitted the identification and mapping of fluvial-channel reservoir fairways and channel belts throughout the study area. In the future, the strategy employed in this study may result in improved well-success rates in the southern Uinta Basin and assist in more detailed reconstructions of the Cedar Mountain and Dakota Formation depositional systems.

With the reality that diverse air interfaces and dissimilar access networks coexist, accompanied by the trend that dynamic spectrum access (DSA) is allowed and will be gradually employed, cognition and cooperation form a promising framework to achieve the ideality of seamless ubiquitous connectivity in future communication networks. In this dissertation, the cognitive gateway (CG), conceived as a special cognitive radio (CR) node, is proposed and designed to facilitate universal interoperability among incompatible waveforms. A proof-of-concept prototype is built and tested. Located in places where various communication nodes and diverse access networks coexist, the CG can be easily set up and works like a network server with differentiated service (Diffserv) architecture to provide automatic traffic relaying and link establishment. The author extracts a scalable '“source-CG-destination“ snapshot from the entire network and investigates the key enabling technologies for such a snapshot. The CG features provide universal interoperability, which is enabled by a generic waveform representation format and the reconfigurable software defined radio platform. According to the trend of an all IP-based solution for future communication systems, the term “waveform“ in this dissertation has been defined as a protocol stack specification suite. The author gives a generic waveform representation format based on the five-layer TCP/IP protocol stack architecture. This format can represent the waveforms used by Ethernet, WiFi, cellular system, P25, cognitive radios etc. A significant advantage of CG over other interoperability solutions lies in its autonomy, which is supported by appropriate signaling processes and automatic waveform identification. The service process in a CG is usually initiated by the users who send requests via their own waveforms. These requests are transmitted during the signaling procedures. The complete operating procedure of a CG is depicted as a waveform-oriented cognition loop, which is primarily executed by the waveform identifier, scenario analyzer, central controller, and waveform converter together. The author details the service process initialized by a primary user (e.g. legacy public safety radio) and that initialized by a secondary user (e.g. CR), and describes the signaling procedures between CG and clients for the accomplishment of CG discovery, user registration and un-registration, link establishment, communication resumption, service termination, route discovery, etc. From the waveforms conveyed during the signaling procedures, the waveform identifier extracts the parameters that can be used for a CG to identify the source waveform and the destination waveform. These parameters are called “waveform indicators.“ The author analyzes the four types of waveforms of interest and outlines the waveform indicators for different types of communication initiators. In particular, a multi-layer waveform identifier is designed for a CG to extract the waveform indicators from the signaling messages. For the physical layer signal recognition, a Universal Classification Synchronization (UCS) system has been invented. UCS is conceived as a self-contained system which can detect, classify, synchronize with a received signal and provide all parameters needed for physical layer demodulation without prior information from the transmitter. Currently, it can accommodate the modulations including AM, FM, FSK, MPSK, QAM and OFDM. The design and implementation details of a UCS have been presented. The designed system has been verified by over-the-air (OTA) experiments and its performance has been evaluated by theoretical analysis and software simulation. UCS can be ported to different platforms and can be applied for various scenarios. An underlying assumption for UCS is that the target signal is transmitted continually. However, it is not the case for a CG since the detection objects of a CG are signaling messages. In order to ensure higher recognition accuracy, signaling efficiency, and lower signaling overhead, the author addresses the key issues for signaling scheme design and their dependence on waveform identification strategy. In a CG, waveform transformation (WT) is the last step of the link establishment process. The resources required for transformation of waveform pairs, together with the application priority, constitute the major factors that determine the link control and scheduling scheme in a CG. The author sorts different WT into five categories and describes the details of implementing the four typical types of WT (including physical layer analog – analog gateway, up to link layer digital – digital gateway, up-to-network-layer digital gateway, and Voice over IP (VoIP) – an up to transport layer gateway) in a practical CG prototype. The issues that include resource management and link scheduling have also been addressed. This dissertation presents a CG prototype implemented on the basis of GNU Radio plus multiple USRPs. In particular, the service process of a CG is modeled as a two-stage tandem queue, where the waveform identifier queues at the first stage can be described as M/D/1/1 models and the waveform converter queue at the second stage can be described as G/M/K/K model. Based on these models, the author derives the theoretical block probability and throughput of a CG. Although the “source-CG-destination” snapshot considers only neighboring nodes which are one-hop away from the CG, it is scalable to form larger networks. CG can work in either ad-hoc or infrastructure mode. Utilizing its capabilities, CG nodes can be placed in different network architectures/topologies to provide auxiliary connectivity. Multi-hop cooperative relaying via CGs will be an interesting research topic deserving further investigation.
Ph. D.

We investigated the relationship between seismo-volcanic events, recorded at La Fossa crater of Vulcano (Aeolian Islands, Italy) during 2004-2009, and the dynamics of the hydrothermal system. During the period of study, six episodes of increasing numbers of seismo-volcanic events took place at the same time as geothermal and geochemical anomalies were observed. These geothermal and geochemical anomalies have been interpreted as resulting from an increasing deep magmatic component of the hydrothermal fluids. Four classes of seismic events (long period, high frequency, monochromatic and tornillos events), characterised by different spectral content and various similarity of the waveforms, have been recognised. These events, clustered mainly below La Fossa crater area at depths of 0.5 1.1 km b.s.l., were space-distributed according to the classes. Based on their features, we can infer that such events at Vulcano are related to two different source mechanisms: (1) fracturing processes of rocks and (2) resonance of cracks (or conduits) filled with hydrothermal fluid. In the light of these source mechanisms, the increase in the number of events, at the same time as geochemical and geothermal anomalies were observed, was interpreted as the result of an increasing magmatic component of the hydrothermal fluids, implying an increase of their flux. Indeed, such variation caused an increase of both the pore pressure within the rocks of the volcanic system and the amount of ascending fluids. Increased pore pressures gave rise to fracturing processes, while the increased fluid flux favoured resonance and vibration processes in cracks and conduits. Finally, a gradual temporal variation of the waveform of the hybrid events (one of the subclasses of long period events) was observed, likely caused by heating and drying of the hydrothermal system. After careful analysis of the seismo-volcanic events of the Aeolian Islands area, the attention was paid to the tectonic events, in order to find possible relationships with the volcanic activity in the area. The aim of this part of the thesis was to identify spatial clusters of earthquakes, locate active seismogenic zone and their relationships with the volcanic activity in the Aeolian Islands. High precision locations were performed in the present thesis, by applying the concept of the velocity model-hypocentres joint inversion and earthquake relocations, along with an analysis of the fault plane solutions. In order to improve our knowledge on the active seismo-tectonics areas we exploited a dataset encompassing 351 events recorded during a 17 year period (1993-2010). Overall, our results show that part of the seismicity is clustered along active seismogenic structures that concur with the main regional tectonic trends whose activity furnishes new elements to better understand the dynamics of the area. A cluster of 24 events in the northern part of Vulcano, NE-SW oriented, marks the presence of a structure that seems to play a key role in magma uprising at Vulcano. These earthquakes suggest the existence of a seismogenic structure (passing just below Vulcanello), which could be interpreted as a discontinuity linking the two magma accumulation zones, thereby representing a possible preferential pathway along which magma may intrude as well as being responsible for fluid migration toward the surface. The results presented in this thesis suggest that the comparison of seismic, ground deformation and temperature data can be useful for better understanding the dynamics of a complex volcano-hydrothermal system, including a better definition of the origin of a volcano unrest, and hence for improving the estimation of the level of the local volcanic hazard.

An essential task for a pacemaker or implantable defibrillator is the accurate identification of rhythm categories so that the correct electrotherapy can be administered. Because some rhythms cause a rapid dangerous drop in cardiac output, it is necessary to categorize depolarization waveforms on a beat-to-beat basis to accomplish rhythm classification as rapidly as possible. In this thesis, a depolarization waveform classifier based on the Lifting Line Wavelet Transform is described. It overcomes problems in existing rate-based event classifiers; namely, (1) they are insensitive to the conduction path of the heart rhythm and (2) they are not robust to pseudo-events. The performance of the Lifting Line Wavelet Transform based classifier is illustrated with representative examples. Although rate based methods of event categorization have served well in implanted devices, these methods suffer in sensitivity and specificity when atrial, and ventricular rates are similar. Human experts differentiate rhythms by morphological features of strip chart electrocardiograms. The wavelet transform is a simple approximation of this human expert analysis function because it correlates distinct morphological features at multiple scales. The accuracy of implanted rhythm determination can then be improved by using human-appreciable time domain features enhanced by time scale decomposition of depolarization waveforms. The purpose of the present work was to determine the feasibility of implementing such a system on a limited-resolution platform. 78 patient recordings were split into equal segments of reference, confirmation, and evaluation sets. Each recording had a sampling rate of 512Hz, and a significant change in rhythm in the recording. The wavelet feature generator implemented in Matlab performs anti-alias pre-filtering, quantization, and threshold-based event detection, to produce indications of events to submit to wavelet transformation. The receiver operating characteristic curve was used to rank the discriminating power of the feature accomplishing dimension reduction. Accuracy was used to confirm the feature choice. Evaluation accuracy was greater than or equal to 95% over the IEGM recordings.

Newly available industry well data and seismic attribute analysis reveal that late Ordovician-early Devonian Hunton Group strata are more widespread (i.e., not removed by mid-Devonian erosion) in the central and southern portions of the Arkoma Basin in eastern Oklahoma than previously thought. This study demonstrates the value of applying seismic attribute analysis to problems of quantifying and mapping stratigraphic features caused by erosions and/or karstification. Well and seismic isochron data in the Red Oak petroleum field for the Viola-Woodford interval (the units that lie stratigraphically beneath and above, respectively, the Huton Group) show isolated ~40-m thick lenses of Hunton rocks, on average measuring 3 km in diameter, with a surrounding halo of karsted rock. This distribution can be explained in two different ways: 1) Hunton occurrences could represent isolated erosional remnants reflecting incomplete removal of the Hunton Group during Middle Devonian time (pre-Woodford unconformity) or 2) due to karsting and collapse of stratigraphically lower units (Viola or Bromide carbonates), lenses of Hunton rocks would have sagged into sinkholes where they were preserved beneath regional base level. Using formation tops from a well data set correlated with attribute and structure maps from a proprietary 3-D seismic data set, we identify three seismic characteristics in the middle Paleozoic interval that correlate well with: 1) absent Hunton seismic markers, indicating that Hunton rocks were completely removed, 2) the Hunton contacts, indicating where a seismically visible section of Hunton rocks remains, 3) absent Hunton but with a thin horizon included within lower carbonate strata that is interpreted to be an incipient karst zone, which is consistently adjacent to areas containing Hunton rocks. The base of the Sylvan Shale and the top of the Woodford Shale, the respective lower and upper adjoining units, form significant chronostratigraphic surfaces. As such, anomalous thicknesses of these units are depositionally related; thick Woodford sections often correlate to thin or absent Hunton rocks, possibly indicating back-filled pre-Woodford channels eroded into or through the Hunton Group. Conversely, when there is little or no Woodford thickening over Hunton lenses and when adjacent areas show thinning and partially karsted Viola rocks, we propose that karstic collapse of Viola strata was responsible for the Hunton rocks preservation. A combination of these models may be necessary to account for areas where we see thinning both in the Woodford and Viola, suggesting that a Hunton lens is structurally lowered due to karsting, but due to its erosionally resistive nature, the lens forms a depositional high, causing the Woodford to thin over it. The 3-D approach is absolutely necessary to reveal the subtle waveform details that illustrate the karstic and erosional processes involved in the preservation of the Hunton wedges. These findings were interpolated, constrained by well data, over the entire Oklahoma portion of the Arkoma basin in order to produce a new Hunton isopach map and 20 separate cross-sections (two shown herein). These show a broad linear region of absent Hunton. Eustatic sea levels rose throughout the middle and late Devonian, so this large area of eroded Hunton is interpreted as a post-Hunton, pre-Woodford structural uplift. Other Hunton wedges, similar in size and extant to that seismically imaged in this study, were also found in the well data. The karstic collapse of the Viola and subsequent preservation of Hunton rocks occurred on both limbs of the arch.

Una tecnica particolarmente importante utilizzata in medicina è il segnale elettroencefalogramma (EEG). La tecnica EEG è stata descritta da Hans Berger, nel 1929, come una "window into the brain" [1]. I segnali EEG sono una registrazione dell'attività elettrica dal cuoio capelluto dei mammiferi, uno strumento fondamentale nella diagnosi e nella ricerca di diversi disturbi cerebrali, comprese quelle relative all'epilessia [2]. Tuttavia, l'analisi delle ore di dati generati dagli EEGs, usato per identificare eventi quali l'epilessia, sonno ed altri è un processo che richiede tempo, in quanto deve essere effettuata da esperti manualmente. Pertanto, per risolvere questo problema, sono stati sviluppati diversi metodi di rilevamento automatico: mimetica, morfologo, template matching, parametric modelling e non-linear features [3]. Nel tentativo di semplificare ulteriormente questa procedura, è stata disegnata una nuova applicazione che vede l'output EEG come un linguaggio, che a sua volta, può essere visto come un codice sorgente di alto livello per computer, e di conseguenza, l'utilizzando di un compilatore per trasformare l'output EEG in una sequenza di simboli. Le proprietà del onda generata potrebbero essere interpretate da un file di grammatica in modo da formare un albero di sintassi astratta, che permettere l'identificazione specifica automatizzata di sotto-tipi di eventi e consentendo l'analisi del contenuto di un evento.
A particularly important technique used in medicine is the electroencephalogram (EEG) signal. EEG technique was described by Hans Berger, in 1929, as a "window into the brain" [1]. EEGs are a recording of electrical activity from the mammalian scalp, a fundamental tool in the diagnosis and research of several brain disorders, including those related to epilepsy [2]. However, the analysis of the hours of data generated from EEGs, used to identify events such as epilepsy, sleep etc, is a time-consuming process, as it has to be performed manually by experts. Therefore, to address this issue, several methods of automatic detection have been developed: mimetic, morphologist, template matching, parametric modelling and non-linear features [3]. In an attempt to further streamline this procedure, a new approach was developed- viewing EEG output as a language, treated as a high level computer source code, and by using a compiler to transform the output into a sequence of symbols, the generated wave’s properties could be interpreted by a grammar file to form an abstract syntax tree. Allowing the specific automated identification of event sub-types and enabling the analysis of the contents of an event.

碩士
國立交通大學
土木工程學系
100
Airborne Lidar is an active remote sensing system. It can obtain the three dimensional coordinates effectively, and provide high density and high precision 3-D point cloud. Full-waveform (FWF) lidar is a new generation of airborne laser scanner which receives one dimensional continuous signal. It offers useful information about the structure of the target. Therefore, the analysis of received signal of FWF lidar and obtaining the implicit information is helpful for landcover classification. In the processing of full waveform Lidar data, the waveform parameter extraction and analysis are the important steps. The major objective of this study is to analyze the received waveform and extract its parameters. We select Gaussian distribution as a symmetric function and Weibull distribution as an asymmetric function in waveform decomposition. Then, we calculate several accuracy assessment indicators between raw waveform data and fitting function for quality assessment. We use echo width, amplitude, backscatter cross-section coefficient, elevation, elevation difference, echo number, and echo ratio as waveform parameter of classification. After waveform parameter extraction, we select Support Vector Machine (SVM) and Random Forests (RF) as classifier for landcover classification. This study employs echo width, amplitude, backscatter cross-section coefficients and other features for classification. Error matrix is used to compare the performance of the classifiers. The experimental results indicate that the accuracy of asymmetric function is slightly better than symmetric function. However, the extracted peak positions from the Gaussian and Weibull are very close. Moreover, Gaussian distribution is relatively simple and easy to implement in the waveform analysis. The result of landcover classification shows that waveform parameters are helpful for classification and Random Forests classifier is better than SVM in our study cases.

碩士
國立成功大學
測量及空間資訊學系碩博士班
101
Thanks to the development of LiDAR technology, recording full waveform information of return laser signal has become available. Compared with the conventional LiDAR system, waveform LiDAR further encodes the intensity of return signal along the time domain, which enables the users to utilize the continuous return signal for the interpretation of ground objects. Potential of more applications than the use of traditional LiDAR can be expected with the use of full waveform LiDAR. A LiDAR waveform is a recorded energy of the backscattered laser pulse along the time domain. The shape of a waveform is formed according as the characteristics surface reflectance, geometric structure and roughness of the laser footprint. It would be possible to extract the information of surface characteristics from waveform data, and this information can be used for the classification of ground surface. This study focuses on the analysis of LiDAR full-waveform data. The effects of various ground objects and surfaces on the waveform data will be analyzed, and the reparability of waveform features among categories of ground objects will be identified. Based on this analysis, a classification approach is developed for LiDAR full-waveform data. The estimation of classification accuracy will be reported as well. The experiment data were collected with three airborne LiDAR systems of different brands, namely Leica、Riegl and Optech. the land cover objects of the experimental area are mainly categorized into road, canopy, grass ＆ crop, bare ground and buildings. Waveform features were analyzed with respect to the single and multiple return laser paths samples, and waveform classification features were selected according to the analysis. Then, the supervised classification by using Support Vector Machine (SVM) and Naive Bayes Classifier (NBC) was performed in three defined methods which include echo-based, waveform-based and waveform-based with images. The experiment results show that the overall accuracy of waveform-based method increases about 20% comparing to echo-based method and it can achieve 86% with the images. This study reveals the potential of 3D object classification using airborne LiDAR waveform data.

碩士
逢甲大學
自動控制工程所
96
In this study, a self-organizing neural network system is presented to classify the real-time electrocardiogram (ECG) signal. This system can not only organize analog waveforms but also output the recognition codes. The system contains a pre-processor and a self-organizing neural network. The pre-processor is to remove the noise in the ECG waveform and segment the data to many samples for the next inputs of the process in the neural network system. The recognized results are useful to find out models of cardiac diseases. By using the signal data in MIT-BIH electrocardiogram database, the proposed neural network system can be trained to create the waveforms of cardiac diseases and to recognize the ECG waveforms. In this research, a novel R point detection method is presented. The test ECG signals include 3,614 R points. The R point detection result by using “So and Chan” method achieves the sensitivity being 76.34%. On the other hand, the proposed R point detection result reaches sensitivity being 99.94%. The proposed method outperforms the previous method. The self-organizing neural network system with MIT-BIH ECG database was tested by creating samples. At first, we segment the 18 data for each with five hours durations to many samples. The RR intervals are segmented from these samples. Following that 5 ECG models (without noise) are created. We also train the self-organizing neural network system with ECG waveform simulator, and obtain 6 ECG models, which have normal and abnormal models. By training the self-organizing neural network system with healthy persons, 3 significant normal ECG models and 3 abnormal ECG models are created. The results of normal and premature ventricular contraction (PVC) samples classifying with MIT-BIH Arrhythmia Database, we separate the samples into normal and abnormal signals by 15 trained normal models, then we classify the abnormal ones by 44 trained PVC models. The accuracy of the classification results is higher than 87%.

博士
國立成功大學
測量及空間資訊學系
103
Compared with discrete LiDAR systems, state-of-the-art airborne waveform LiDAR systems provide richer information on illuminated surfaces. Waveform data contains both the spatial and physical information of the surfaces. The geospatial surfaces can be located by detecting the reflected laser signal stored in the waveform with the information of the laser travelling path. The process to detect the reflected signal is known as echo detection. The physical characteristics of surfaces such as the reflectance or surface roughness will deform the shape of the transmitting laser pulse resulting in different waveform features. Such features can be used for land cover classification. For waveform information extraction, the echoes are usually detected before the waveform features are extracted for further analysis. For echo detection, conventional discrete LiDAR systems often use an on-the-fly process to detect points. This process usually misdetects weak or overlapping echoes, thus resulting in poor geometry when the structure of a scanned area is complex, such as a forest area. This study proposes an echo detection approach based on wavelet transformation that is capable of detecting weak returns and resolving overlapping echoes. Simulated and real waveform datasets of a forest area were both used in this study. The simulated waveform data were utilized to compare the proposed detector with two other popular detectors, namely, zero crossing (ZC) and Gaussian decomposition (GD), in terms of their ability to deal with weak or overlapping echoes. The real waveform dataset were used to demonstrate the wavelet-based (WB) algorithm for exploring missing echoes. Experiments using simulated data showed that the WB and GD detectors are superior to the ZC detector in finding overlapping echoes. The WB algorithm performs well when dealing with overlapping echoes with low signal-to-noise ratio. The proposed WB algorithm was then applied to the real waveform dataset to test its effectiveness in detecting missing echoes. Results show that the WB algorithm can find more than 31.5% number of points than that of the used LiDAR system. An automatic filtering process was applied to the point clouds extracted from the waveform data to classify the ground points. This paper presents assessment methods based on the visual analyses of point classification and on the elevation difference of generated digital elevation models. Results show that the filtering accuracy and the accuracy of the digital elevation model are both improved because an enhanced geometry of the landscape can be obtained from the detected points. For land cover classification, features that can be derived from waveform data to describe land covers are divided into two categories, namely, echo-based and waveform-based features. Echo-based features have been widely used by previous studies to effectively classify land covers when the waveform has a single return. When the waveform contains multi-returns, echo-based features would fail to distinguish some land covers. Thus, waveform-based features are used and investigated in this study to complement the disadvantages of echo-based features. Experiments show that land cover classification can be improved with the integration of echo-based and waveform-based features.

This thesis work consists of two research projects. The first project presented is on waveform design for car radars. These radars are used to detect other vehicles to avoid collision. In this project, we attempt to find the best waveform that distinguishes large objects from small ones. This helps the radar system reach more reliable decisions. We consider several models of the problem with varying complexity. For each model, we present optimization results calculated under various constraints regarding how the waveform is generated and how the reflected signal is processed. The results show that changing the radar waveform can result in better target classification. The second project is about digital-to-analog converter (DAC) linearization. Ideally, DACs have a linear input-output relation. In practice, however, this relation is nonlinear which may be harmful for many applications. A more linear input-output relation can be achieved by modifying the input to a DAC. This method, called predistortion, requires a good understanding of how DAC errors contribute to the nonlinearity. Assuming a simple DAC model, we investigate how different error functions lead to different types of nonlinearities through theoretical analyses and supporting computer simulations. We present our results in terms of frequency spectrum calculations. We show that the nonlinearity observed at the output strongly depends on how the error is modeled. These results are helpful in designing a predistorter for linearization.

碩士
國立臺灣大學
電信工程學研究所
102
Artificial neural network is a kind of machine learning tools; it’s a simplified model of the brain, imitating biological neural networks. The human brain is able to learn from experience, and has good performance on visual and audio signal processing. By imitating human brain neural networks, people expect to bring computers the same ability as human that can help people to solve various problems. Combined with neuroscience knowledge, a more physiological meaningful tool, spiking neural network, has been created. The spiking neural network transmits information by spike trains and imitates the membrane potential function of neurons. Hence spiking neural network has the better performance on classification and prediction, and the work function is more similar to the human brain. Now spiking neural network is used for neuroscience simulation and machine learning application. However, there are few machine learning applications of spiking neural network. This study designed a spiking neural network based waveform classification structure with an application of arrhythmia pattern recognition. There was discussion of encoding methods and functionality of spiking neurons. Furthermore, we modified the Tempotron algorithm to improve the accuracy of prediction. At last, we got the good performance in the tests of MIT-BIH arrhythmia database and NTUH telehealth database. This study proposed a new application of spiking neural networks and proved the ability and potential of spiking neural networks.

碩士
國防大學理工學院
空間科學碩士班
102
LiDAR (Light Detection And Ranging), or called Airborne Laser Scannimg, is an active telemetry to surface. It can quickly obtain high density and precision threedimensional coordinates by using direct geographical position, Along with the development of technology, the current LiDAR system of business model can completely record its reflection waveform. The implied information, such as the reflective physical characteristics and details of changes in surface features. It compared to traditional discrete LiDAR system, full waveform LiDAR system provides users with more information for research. The system now has become a large range of high-density and high-precision three-dimensional surface information of the technology. It used waveform LiDAR data with multi-strips and different flight mission in Taichung urban area acquired from cloud points of the waveform features: width, amplitude and backscatter cross-section coefficient in this study. It analysed the waveform features with different factor of the scan time, scan date and flight altitude in experiment because each strip with different scanning conditions. It selected the optimum features for landcover classification was: the standard deviation of width and the mean of backscatter cross-section coefficient, and added the geometric features with echo ratio and normalized digital surface model and other. According to the landcover features of buildings, asphalt, cement road and vegetation to feature analysis and landcover classification. Classification outcomes showed that used threshold of waveform features after landcover feature analysis to the multi-strips and different missions. Its overall accuracy can reach more than 80%, Kappa value could reach 0.70 or more.

碩士
國立中央大學
土木工程學系
102
Airborne Full-Waveform LiDAR (FW) is an active remote sensing system. It not only provides the three-dimensional coordinates about the ground objects but also record the whole return signal as the waveform. The physical properties of objects in a ray can be obtained by fitting and analyzing the waveform. It offers useful information to user for three dimensional reconstruction and land cover distinguishing. In the processing of FW LiDAR data for land-cover classification, the waveform fitting and analysis are the first steps. In this study, the waveform data was fitted by cubic smoothing spline after eliminating the background noise. The amplitude and width was derived based on the peaks detected by second derivative method. In the case of the waveform with the multiple returns, the feature such as time difference of first and last return, peak numbers, and average amplitude was obtained. These waveform parameters combined with intensity and normalized height was utilized as the features for land-cover classification. The classifier used in this study was Random Forest. In order to discuss the effect of cubic smoothing spline, the classification result was compared to the Gaussian decomposition method which is a popular method in full-waveform application. The experimental results indicate that cubic smoothing spline provide the smaller fitting error and keep more information in the waveform. The land cover classification results demonstrate that the multiple return features are helpful for the building edge and trees which are easily misclassified. In addition, cubic smoothing spline is suitable for full-waveform Lidar data with the better classification result and efficiency.

Wissenschaftliche Untersuchungen von terrestrischen und aquatischen Ökosystemen erfordern präzise Informationen über die dreidimensionale Struktur des ökologischen Systems. Full-waveform Flugzeuglaserscannerdaten eignen sich hervorragend zur Charakterisierung von Ökosystemen und bilden eine ideale Basis für die vollständige volumetrische Repräsentation der Vegetations- und Gewässerstruktur in einem Voxelraum. Die Voxelattribute werden dabei aus der digitalisierten Wellenform abgeleitet. Jeder emittierte Laserpuls wird von Dämpfungseffekten beeinflusst, die durch Teilreflexionen auf seinem Weg durch die unterschiedlichen Vegetations- oder Wasserschichten entstehen. Dadurch ist die Struktur im unteren Bereich der empfangenen Rohsignale unterrepräsentiert. Die im Rahmen dieser Arbeit entwickelten innovativen Methoden zur Analyse von full-waveform Daten ermöglichen die Generierung einer radiometrisch korrigierten Voxelraumrepräsentation. Voraussetzung dafür ist die numerisch stabile Rekonstruktion des effektiven differentiellen Rückstreuquerschnitts mit geeigneten Entfaltungs- und Regularisierungsverfahren. Das Kernstück der Analyse bildet die Beschreibung der Signaldämpfung mit Hilfe geeigneter Modelle. Auf Grundlage dieser Modelle wurden neuartige Korrekturverfahren zur Kompensation der Signaldämpfung erarbeitet, wobei der Korrekturterm direkt aus dem differentiellen Rückstreuquerschnitt abgeleitet wird. Die Grundidee der entwickelten Methode ist das schrittweise Anheben der Signalintensität in Abhängigkeit von der individuellen Historie jedes Laserpulses. Die Resultate der vorliegenden Arbeit tragen dazu bei, die in full-waveform Daten enthaltenen Informationen über die Vegetations- und Gewässerstruktur zugänglich zu machen. Weiterhin zeigen die hier präsentierten Ergebnisse, dass die Limitierungen bestehender Auswertemethoden, welche weitgehend auf die Extraktion diskreter Maxima und die Erzeugung volumetrischer Repräsentationen aus diskreten 3D Punktwolken beschränkt sind, überwunden werden können.
The scientific investigation of terrestrial and aquatic ecosystems requires precise information on the three-dimensional structure of the ecologic system. Full-waveform airborne laser scanner data are an ideal basis for the complete volumetric representation of vegetation and water structure in a voxel space. Due to attenuation effects, caused by partial reflections during the laser pulse propagation through the vegetation or water column, each individual laser pulse echo is significantly modified. As a result, the structure in the lower parts of the vegetation or water column is underrepresented in the digitized waveform. Within this research, novel and innovative methods were developed, which enable the generation of a radiometrically correct voxel space representation. Therefore, a numerically stable reconstruction of the effective differential backscattering cross section utilizing appropriate deconvolution and regularization techniques is required. The essential element of the analysis is the description of the signal attenuation using applicable mathematical models. For this purpose, novel correction methods compensating the signal attenuation based on these models were developed. The correction term is directly derived from the differential backscatter cross section. The basic idea is a gradually increase of the signal amplitudes depending on the individual history of each laser pulse. The results gained in this work contribute to an improved access to the information on vegetation and water structure, contained in full-waveform laser scanner data. Furthermore, it is possible to overcome limitations of existing approaches, which are mainly based on the extraction of discrete maxima.

碩士
國立中央大學
土木工程學系
101
Airborne LiDAR is an active remote sensing system. It can transfer the distance and orientation to point cloud by direct georeferencing. A new generation of LiDAR system called Full-Waveform (FW) LiDAR which could receive the whole return signal (so-called waveform) in a ray has become popular recently. With FW LiDAR, users can obtain more information of objects by analyzing the waveform, and it is helpful for three dimensional reconstruction and land cover distinguishing. In the processing of FW LiDAR data, the waveform parameter extraction and analysis are the important steps. In this study, after eliminating the background noise, the Gaussian modeling function with second derivative method was used for waveform fitting. The result was compared to Gaussian fitting using the initial value provided by the instrument. Then the features extracted from the waveform, including width, amplitude, backscatter cross-section, traditional LiDAR features, normalized height and intensity, and greenness index from image were used for land cover classification. The classifiers used in this study were Naïve Bayes and Random Forest and compared with each other. The experimental results indicate that using the second derivative method could provide higher fitting successful rate, smaller Root Mean Square Error (RMSE) and better classification result. The land cover classification results demonstrate that full-waveform features are helpful for distinguishing different vegetation targets and the decision-tree-based Random Forest classifier is more suitable for landcover classification of LiDAR data used in this study.

碩士
國立成功大學
測量及空間資訊學系
105
Satellite radar altimetry becomes an irreplaceable tool to provide accurate surface height measurements over open oceans. However, the accuracy decreases when altimeters approach coastlines or non-ocean surfaces due to the improper geophysical corrections and complex returned waveforms. Many waveform retracking algorithms have been developed for improving the accuracy of non-ocean reflected altimetry data; however, the performance still cannot achieve the same accuracy as that in open oceans. In coastal regions, some waveforms reflected from non-ocean surfaces lead to the worse retracking results. Therefore, waveform classification methods are needed to distinguish if waveforms are truly reflected from oceans. Waveform classification used in this study includes two steps. The first step is applying Linear Discriminant Analysis (LDA) to reduce the dimensionality of original features’ spaces for classification. The second step is using k-Nearest Neighbors Classifier (k-NN) to separate waveforms into two groups: ocean and non-ocean waveforms. Afterward, we remove the non-ocean waveform before doing retracking. In this study, we use Envisat altimetry data over Hsiang-Shan wetland in Hsin-Chu, which is located in Northwestern Taiwan. The satellite-derived results are then evaluated using Hsin-Chu tide gauge data. In the case of distance from coastline 0~5 km, after waveform classification, the best performance retracker is ice-1 and standard deviation of the difference between tide gauge and ice-1 improve from 1.140 m to 0.173 m. Finally, we expect building an effective classification method and figuring out the most appropriate retracking algorithm applied for this study area.

abstract: Genomic and proteomic sequences, which are in the form of deoxyribonucleic acid (DNA) and amino acids respectively, play a vital role in the structure, function and diversity of every living cell. As a result, various genomic and proteomic sequence processing methods have been proposed from diverse disciplines, including biology, chemistry, physics, computer science and electrical engineering. In particular, signal processing techniques were applied to the problems of sequence querying and alignment, that compare and classify regions of similarity in the sequences based on their composition. However, although current approaches obtain results that can be attributed to key biological properties, they require pre-processing and lack robustness to sequence repetitions. In addition, these approaches do not provide much support for efficiently querying sub-sequences, a process that is essential for tracking localized database matches. In this work, a query-based alignment method for biological sequences that maps sequences to time-domain waveforms before processing the waveforms for alignment in the time-frequency plane is first proposed. The mapping uses waveforms, such as time-domain Gaussian functions, with unique sequence representations in the time-frequency plane. The proposed alignment method employs a robust querying algorithm that utilizes a time-frequency signal expansion whose basis function is matched to the basic waveform in the mapped sequences. The resulting WAVEQuery approach is demonstrated for both DNA and protein sequences using the matching pursuit decomposition as the signal basis expansion. The alignment localization of WAVEQuery is specifically evaluated over repetitive database segments, and operable in real-time without pre-processing. It is demonstrated that WAVEQuery significantly outperforms the biological sequence alignment method BLAST for queries with repetitive segments for DNA sequences. A generalized version of the WAVEQuery approach with the metaplectic transform is also described for protein sequence structure prediction. For protein alignment, it is often necessary to not only compare the one-dimensional (1-D) primary sequence structure but also the secondary and tertiary three-dimensional (3-D) space structures. This is done after considering the conformations in the 3-D space due to the degrees of freedom of these structures. As a result, a novel directionality based 3-D waveform mapping for the 3-D protein structures is also proposed and it is used to compare protein structures using a matched filter approach. By incorporating a 3-D time axis, a highly-localized Gaussian-windowed chirp waveform is defined, and the amino acid information is mapped to the chirp parameters that are then directly used to obtain directionality in the 3-D space. This mapping is unique in that additional characteristic protein information such as hydrophobicity, that relates the sequence with the structure, can be added as another representation parameter. The additional parameter helps tracking similarities over local segments of the structure, this enabling classification of distantly related proteins which have partial structural similarities. This approach is successfully tested for pairwise alignments over full length structures, alignments over multiple structures to form a phylogenetic trees, and also alignments over local segments. Also, basic classification over protein structural classes using directional descriptors for the protein structure is performed.
Dissertation/Thesis
Ph.D. Electrical Engineering 2011

Η παρακολούθηση του ρεύματος διαρροής, και ειδικά της κυματομορφής του, είναι μια ευρύτατα διαδεδομένη τεχνική για την παρακολούθηση της επιφανειακής δραστηριότητας και κατάστασης των μονωτήρων υψηλής τάσης. Η παρακολούθηση στο πεδίο είναι απαραίτητη για να υπάρξει μια πιστή καταγραφή της δραστηριότητας και συμπεριφοράς σε πραγματικές συνθήκες, παρουσιάζει όμως σημαντικές δυσκολίες. Το πρόβλημα συνήθως παρακάμπτεται με την καταγραφή και μελέτη εξαγόμενων μεγεθών όπως η τιμή κορυφής και το φορτίο, μία προσέγγιση που οδηγεί όμως σε αμφίβολα αποτελέσματα. Η παρούσα διατριβή επικεντρώνεται στην διερεύνηση και ταξινόμηση κυματομορφών ρεύματος διαρροής καταγεγραμμένων στο πεδίο. Αρχικά, παρατίθεται μια λεπτομερής ανασκόπηση της καταγραφής και ανάλυσης ρεύματος διαρροής σε εργαστηριακές και πραγματικές συνθήκες. Στην συνέχεια, περιγράφεται το πεδίο μετρήσεων, δύο Υποσταθμοί Υψηλής Τάσης 150kV, το αναπτυχθέν λογισμικό αλλά και ο Υπαίθριος Σταθμός Δοκιμών όπου πρόκειται να αξιοποιηθούν τα αποτελέσματα. Μελετώνται περισσότερες από 100.000 κυματομορφές, που έχουν καταγραφεί σε μια περίοδο που ξεπερνάει τα δέκα έτη. Εξετάζεται και αξιολογείται το πρόβλημα του θορύβου και ταυτοποιούνται τρεις διαφορετικοί τύποι θορύβου. Εξετάζεται η επίδρασή τους στο πρόβλημα συσσώρευσης δεδομένων αλλά και στην ποιότητα της εξαγόμενης πληροφορίας. Για την αντιμετώπιση του προβλήματος εφαρμόζονται και αξιολογούνται τρεις διαφορετικές τεχνικές. Για την περαιτέρω ταξινόμησή των κυματομορφών που απεικονίζουν δραστηριότητα, χρησιμοποιούνται διάφορες τεχνικές επεξεργασίας σήματος, εξαγωγής και επιλογής χαρακτηριστικών καθώς και αναγνώρισης προτύπων όπως η Wavelet Multi-Resolution Ανάλυση, η Ανάλυση Fourier, τα Νευρωνικά Δίκτυα, το t-test, ο αλγόριθμος mRMR, ο αλγόριθμος κ-πλησιέστερων γειτόνων, ο απλός Μπεϋζιανός ταξινομητής και οι Μηχανές Διανυσμάτων Υποστήριξης. Συγκεντρωτικά, δίνεται μια συνολική εικόνα των διαφορετικών ζητημάτων που σχετίζονται με την παρακολούθηση του ρεύματος διαρροής. Παρατίθεται μια πλήρης εικόνα των κυματομορφών όπως αυτές καταγράφονται σε πραγματικές συνθήκες, υπογραμμίζοντας ιδιαιτερότητες που σχετίζονται με την φύση της εφαρμογής. Εφαρμόζονται και αξιολογούνται νέες προσεγγίσεις για την ταξινόμηση των κυματομορφών. Τα συνολικά αποτελέσματα προσφέρουν σημαντική ενίσχυση στην αποτελεσματικότητα της τεχνικής της παρακολούθησης του ρεύματος διαρροής, συμβάλλοντας σημαντικά στην μελέτη της επιφανειακής δραστηριότητας και συμπεριφοράς των μονωτήρων υψηλής τάσης.
Leakage current monitoring is a widely applied technique for monitoring surface activity and condition of high voltage insulators. Field monitoring is necessary to acquire an exact image of activity and performance in the field. However, recording, managing and interpreting leakage current waveforms, the shape of which is correlated to surface activity, is a major task. The problem is commonly by-passed with the extraction, recording and investigation of values related to peak and charge, an approach reported to produce questionable results. The present thesis focuses on the investigation and classification of field leakage current waveforms. At first, a detailed background of measuring and analyzing leakage current both in lab and field conditions is provided. Then, the monitoring sites, two 150kV Substations, as well as the developed custom-made software and the newly constructed High Voltage Test Station where the results of this thesis are to be implemented, is briefly described. More than 100.000 waveforms are investigated, recorded through a period exceeding ten years. Field related noise is thoroughly described and evaluated. Three different types of noise are identified and their impact on the size of accumulated data and on data interpretation is investigated. Three different techniques to overcome the problem are applied and evaluated. Activity portraying waveforms are further investigated. Further classification of activity portraying waveforms is performed employing signal processing, feature extraction and selection algorithms as well as pattern recognition techniques such as Wavelet Multi-Resolution Analysis, Fourier Analysis, Neural Networks (NNs), student’s t-test, minimum Redundancy Maximum Relevance (mRMR), k-Nearest Neighbors (kNN), Naive Bayesian Classifier and Support Vector Machines (SVMs). Overall results provide a full image of the various aspects of field leakage current monitoring. A detailed image of field waveforms, revealing several new attributes, is documented. New approaches for the classification of leakage current waveforms are introduced, applied on field waveforms and evaluated. Results described in this thesis significantly enhance the effectiveness of the leakage current monitoring technique, providing a powerful tool for the investigation of surface activity and performance of high voltage insulators.

This paper proposes a cumulant based classification means to identify waveforms for a blind receiver in the presence of time varying channels, which is built from the work done on cumulants in static channels currently in the literature. Results show the classification accuracy is on the order or better than current methods in use in static channels that do not vary over an observation period. This is accomplished by making use of second through tenth order cumulants in a signature vector that the search engine platform has the means of differentiating. A receiver can then blindly identify waveforms accurately in the presence of multipath Rayleigh fading with AWGN noise. Channel learning occurs prior to classification in order to identify the consistent distortion pattern for a waveform that is observable in the signature vector. Then using a database look-up method, the observed waveform is identified as belonging to a particular cluster based on the observed signature vector. If the distortion patterns are collected from a variety of channel types, the database can then classify both the waveform and the rough channel type that the waveform passed through. If the exact channel model or channel parameters is known and used as a limiter, significant improvement on the waveform classification can be achieved. Greater accuracy comes from using the exact channel model as the limiter.
Master of Science

Software Audiosynthesizer haben in den letzten 10 Jahren enorm an Popularität gewonnen und sind in vielen Profi- und Heimstudios nicht mehr wegzudenken. Diese Popularität ist durch die hohe Rechenleistung begründet, welche auf PCs und mobilen Geräten überall zur Verfügung steht und Echtzeitaudiosynthese nutzbar macht. Das Ziel dieser Arbeit ist die ausührliche Beschreibung grundlegender Synthesizerkomponenten und die Untersuchung geeigneter Algorithmen und Techniken für deren Realisierung.

The advancements in wireless communications are the key-enablers of new applications with stringent requirements in low-latency, ultra-reliability, high data rate, high mobility, and massive connectivity. Diverse types of devices, ranging from tiny sensors to vehicles, with different capabilities need to be connected under various channel conditions. Thus, modern connectivity and network techniques at all layers are essential to overcome these challenges. In particular, the physical layer (PHY) transmission is required to achieve certain link reliability, data rate, and latency. In modern digital communications systems, the transmission is performed by means of a digital signal processing module that derives analog hardware. The performance of the analog part is influenced by the quality of the hardware and the baseband signal denoted as waveform. In most of the modern systems such as fifth generation (5G) and WiFi, orthogonal frequency division multiplexing (OFDM) is adopted as a favorite waveform due to its low-complexity advantages in terms of signal processing. However, OFDM requires strict requirements on hardware quality. Many devices are equipped with simplified analog hardware to reduce the cost. In this case, OFDM does not work properly as a result of its high peak-to-average power ratio (PAPR) and sensitivity to synchronization errors. To tackle these problems, many waveforms design have been recently proposed in the literature. Some of these designs are modified versions of OFDM or based on conventional single subcarrier. Moreover, multicarrier frameworks, such as generalized frequency division multiplexing (GFDM), have been proposed to realize varieties of conventional waveforms. Furthermore, recent studies show the potential of using non-conventional waveforms for increasing the link reliability with affordable complexity. Based on that, flexible waveforms and transmission techniques are necessary to adapt the system for different hardware and channel constraints in order to fulfill the applications requirements while optimizing the resources. The objective of this thesis is to provide a holistic view of waveforms and the related multiple access (MA) techniques to enable efficient study and evaluation of different approaches. First, the wireless communications system is reviewed with specific focus on the impact of hardware impairments and the wireless channel on the waveform design. Then, generalized model of waveforms and MA are presented highlighting various special cases. Finally, this work introduces low-complexity architectures for hardware implementation of flexible waveforms. Integrating such designs with software-defined radio (SDR) contributes to the development of practical real-time flexible PHY.:1 Introduction 1.1 Baseband transmission model 1.2 History of multicarrier systems 1.3 The state-of-the-art waveforms 1.4 Prior works related to GFDM 1.5 Objective and contributions 2 Fundamentals of Wireless Communications 2.1 Wireless communications system 2.2 RF transceiver 2.2.1 Digital-analogue conversion 2.2.2 QAM modulation 2.2.3 Effective channel 2.2.4 Hardware impairments 2.3 Waveform aspects 2.3.1 Single-carrier waveform 2.3.2 Multicarrier waveform 2.3.3 MIMO-Waveforms 2.3.4 Waveform performance metrics 2.4 Wireless Channel 2.4.1 Line-of-sight propagation 2.4.2 Multi path and fading process 2.4.3 General baseband statistical channel model 2.4.4 MIMO channel 2.5 Summary 3 Generic Block-based Waveforms 3.1 Block-based waveform formulation 3.1.1 Variable-rate multicarrier 3.1.2 General block-based multicarrier model 3.2 Waveform processing techniques 3.2.1 Linear and circular filtering 3.2.2 Windowing 3.3 Structured representation 3.3.1 Modulator 3.3.2 Demodulator 3.3.3 MIMO Waveform processing 3.4 Detection 3.4.1 Maximum-likelihood detection 3.4.2 Linear detection 3.4.3 Iterative Detection 3.4.4 Numerical example and insights 3.5 Summary 4 Generic Multiple Access Schemes 57 4.1 Basic multiple access and multiplexing schemes 4.1.1 Infrastructure network system model 4.1.2 Duplex schemes 4.1.3 Common multiplexing and multiple access schemes 4.2 General multicarrier-based multiple access 4.2.1 Design with fixed set of pulses 4.2.2 Computational model 4.2.3 Asynchronous multiple access 4.3 Summary 5 Time-Frequency Analyses of Multicarrier 5.1 General time-frequency representation 5.1.1 Block representation 5.1.2 Relation to Zak transform 5.2 Time-frequency spreading 5.3 Time-frequency block in LTV channel 5.3.1 Subcarrier and subsymbol numerology 5.3.2 Processing based on the time-domain signal 5.3.3 Processing based on the frequency-domain signal 5.3.4 Unified signal model 5.4 summary 6 Generalized waveforms based on time-frequency shifts 6.1 General time-frequency shift 6.1.1 Time-frequency shift design 6.1.2 Relation between the shifted pulses 6.2 Time-frequency shift in Gabor frame 6.2.1 Conventional GFDM 6.3 GFDM modulation 6.3.1 Filter bank representation 6.3.2 Block representation 6.3.3 GFDM matrix structure 6.3.4 GFDM demodulator 6.3.5 Alternative interpretation of GFDM 6.3.6 Orthogonal modulation and GFDM spreading 6.4 Summary 7 Modulation Framework: Architectures and Applications 7.1 Modem architectures 7.1.1 General modulation matrix structure 7.1.2 Run-time flexibility 7.1.3 Generic GFDM-based architecture 7.1.4 Flexible parallel multiplications architecture 7.1.5 MIMO waveform architecture 7.2 Extended GFDM framework 7.2.1 Architectures complexity and flexibility analysis 7.2.2 Number of multiplications 7.2.3 Hardware analysis 7.3 Applications of the extended GFDM framework 7.3.1 Generalized FDMA 7.3.2 Enchantment of OFDM system 7.4 Summary 7 Conclusions and Future works