Detection and Classification of Whale Acoustic Signals

This dissertation focuses on two vital challenges in relation to whale acoustic signals: detection and classification.

In detection, we evaluated the influence of the uncertain ocean environment on the spectrogram-based detector, and derived the likelihood ratio of the proposed Short Time Fourier Transform detector. Experimental results showed that the proposed detector outperforms detectors based on the spectrogram. The proposed detector is more sensitive to environmental changes because it includes phase information.

In classification, our focus is on finding a robust and sparse representation of whale vocalizations. Because whale vocalizations can be modeled as polynomial phase signals, we can represent the whale calls by their polynomial phase coefficients. In this dissertation, we used the Weyl transform to capture chirp rate information, and used a two dimensional feature set to represent whale vocalizations globally. Experimental results showed that our Weyl feature set outperforms chirplet coefficients and MFCC (Mel Frequency Cepstral Coefficients) when applied to our collected data.

Since whale vocalizations can be represented by polynomial phase coefficients, it is plausible that the signals lie on a manifold parameterized by these coefficients. We also studied the intrinsic structure of high dimensional whale data by exploiting its geometry. Experimental results showed that nonlinear mappings such as Laplacian Eigenmap and ISOMAP outperform linear mappings such as PCA and MDS, suggesting that the whale acoustic data is nonlinear.

We also explored deep learning algorithms on whale acoustic data. We built each layer as convolutions with either a PCA filter bank (PCANet) or a DCT filter bank (DCTNet). With the DCT filter bank, each layer has different a time-frequency scale representation, and from this, one can extract different physical information. Experimental results showed that our PCANet and DCTNet achieve high classification rate on the whale vocalization data set. The word error rate of the DCTNet feature is similar to the MFSC in speech recognition tasks, suggesting that the convolutional network is able to reveal acoustic content of speech signals.

(Table 1) Radiocarbon ages of core AB01/94 obtained in the Tromper Wiek, western Baltic Sea

In the Tromper Wiek northeast of Rügen, acoustical investigations using Air Gun, Boomer, Chirp Sonar and Sediment-Echosounder were carried out. Together with sediment core information, it allowed the identification of five seismostratigraphic units (E1 to E5). Conventional and AMS-14C-datings supported their chronostratigraphical classification. The uppermost till (E1) was incised by late glacial channels filled with glaciolacustrine sediments (E2) of the early Baltic Ice Lake stages. These were regionally overlain with a sharp unconformity by a thick (locally >20 m) sedimentary complex (E3) of acoustically laminated silts of freshwater origin. This lower part of the E3-complex (E3a) is overlain by fluvial to coastal silty fine sands (E3b) deeper towards the Arkona Basin. Fine plant debris in the uppermost part of sub-unit E3a yielded ages of 10,100 and 10,500 14C-years B.P., representing the final phase of the Baltic Ice Lake. The fine sands of sub-unit E3b were deposited after the final drainage of the Baltic Ice Lake. In the shallower central part of the bay, the silts of sub-unit E3a were covered by a younger unit (E4) of fine sand with plant debris. A sedge peat occurring at the basis of unit E4 yielded an age of 9,590 14C-years B.P. The fine sands overlying the unit E3 in the central part of Tromper Wiek were deposited in the Ancylus Lake. Their position at about 20 m below present sea level (b.s.l.) reflects the maximum highstand in this area. The character and distribution of the Early Holocene deposits at greater depth suggest a lake water level at about 30 m b.s.l. after this highstand. Below 25 m b.s.l. muddy Littorina Sea sediments are observed. The thickness of these muds and sandy muds increases gradually towards the Arkona Basin. Locally, they are found in a channel-like structure immediately north of Jasmund.

L'impact du glissement en fréquence lors de l'accélération directe d'électrons par le faisceau laser

L’accélération directe d’électrons par des impulsions ultrabrèves de polarisation radiale fortement focalisées démontre un grand potentiel, notamment, pour la production de paquets d’électrons ultrabrefs. Plusieurs aspects de ce schéma d’accélération restent toutefois à être explorés pour en permettre une maîtrise approfondie. Dans le cadre du présent mémoire, on s’intéresse à l’ajout d’une dérive de fréquence au champ de l’impulsion TM01 utilisée. Les expressions exactes des composantes du champ électromagnétique de l’impulsion TM01 sont établies à partir d’une généralisation du spectre de Poisson. Il s’agit, à notre connaissance, du premier modèle analytique exact pour la description d’une impulsion avec une dérive de fréquence. Ce modèle est utilisé pour étudier l’impact du glissement en fréquence sur le schéma d’accélération, grâce à des simulations “particule test” unidimensionnelles, considérant en premier lieu une énergie constante par impulsion, puis un champ maximum constant. Les résultats révèlent que le glissement en fréquence diminue le gain en énergie maximum atteignable dans le cadre du schéma d’accélération à l’étude ; une baisse d’efficacité de plusieurs dizaines de pourcents peut survenir. De plus, les simulations mettent en évidence certaines différences reliées à l’utilisation d’impulsions avec une dérive vers les basses fréquences ou avec une dérive vers les hautes fréquences : il se trouve que, pour un glissement en fréquence de même grandeur, l’impulsion avec une dérive vers les basses fréquences conduit à un gain en énergie cinétique maximum plus élevé pour l’électron que l’impulsion avec une dérive vers les hautes fréquences.

Digital predistortion for compensation of nonlinearities in Radio over Fiber Links

In order to cope up with the ever increasing demand for larger transmission bandwidth, Radio over Fiber technology is a very beneficial solution. These systems are expected to play a major role within future fifth generation wireless networks due to their inherent capillary distribution properties. Nonlinear compensation techniques are becoming increasingly important to improve the performance of telecommunication channels by compensating for channel nonlinearities. Indeed, significant bounds on the technology usability and performance degradation occur due to nonlinear characteristics of optical transmitter, nonlinear generation of spurious frequencies, which, in the case of RoF links exploiting Directly Modulated Lasers , has the combined effect of laser chirp and optical fiber dispersion among its prevailing causes. The purpose of the research is to analyze some of the main causes of harmonic and intermodulation distortion present in Radio over Fiber (RoF) links, and to suggest a solution to reduce their effects, through a digital predistortion technique. Predistortion is an effective and interesting solution to linearize and this allows to demonstrate that the laser’s chirp and the optical fiber’s dispersion are the main causes which generate harmonic distortion. The improvements illustrated are only theoretical, based on a feasibility point of view. The simulations performed lead to significant improvements for short and long distances of radio over fiber link lengths. The algorithm utilized for simulation has been implemented on MATLAB. The effects of chirp and fiber nonlinearity in a directly modulated fiber transmission system are investigated by simulation, and a cost effective and rather simple technique for compensating these effects is discussed. A detailed description of its functional model is given, and its attractive features both in terms of quality improvement of the received signal, and cost effectiveness of the system are illustrated.

Caracterização geológica e geofísica da área do porto da Figueira da Foz

No âmbito de um projeto realizado para a Administração Portuária do Porto da Figueira da Foz, foram recolhidos dados geofísicos no offshore da costa da Figueira da Foz, utilizando um sistema combinado de sonar de varrimento lateral e de sísmica de alta resolução com o objetivo de compreender a distribuição de sedimentos ao longo do fundo marinho e assim entender melhor a sua origem. Os dados de varrimento lateral foram processados, analisados e representados em mosaicos recorrendo ao software integrado TRINTON Prespective. Foram colhidas 10 amostras de sedimentos de fundo que foram analisadas do ponto de vista granulométrico e que foram usadas para a calibração dos mosaicos do sonar de varrimento lateral de modo a produzir um mapa semi-quantitativo de classificação dos sedimentos de fundo. Os dados de reflexão sísmica foram processados e analisados através dos softwares SPW e RadExPro. Os mapas de sonar de varrimento lateral produzidos (mosaicos) mostram uma relação entre o tamanho de grão e composição dos depósitos de plataforma interna e os processos geológicos que ocorrem na zona costeira adjacente. Também se verifica uma relação entre a orientação da costa e a sua consequente exposição às ondas incidentes, e como isso afeta e altera a distribuição do tamanho grão. Neste contexto, a interpretação combinada dos mosaicos do sonar de varrimento lateral com os dados de granulometria e a batimetria dos sedimentos permitiram uma cartografia interpretativa detalhada da morfologia do mar, uma melhor compreensão do controlo da linha costeira na propagação das ondas e correntes e, por conseguinte, na transferência de sedimento no litoral. Os dados de sísmica permitem ainda verificar a espessura das várias camadas geológicas em profundidade.

Stratigraphic Architecture of Deep Sea Depositional Systems in the Southern Tyrrhenian Sea: Some Examples in the Ischia and Stromboli Volcanic Islands (Southern Italy)

The stratigraphic architecture of deep sea depositional systems has been discussed in detail. Some examples in Ischia and Stromboli volcanic islands (Southern Tyrrhenian sea, Italy) are here shown and discussed. The submarine slope and base of slope depositional systems represent a major component of marine and lacustrine basin fills, constituting primary targets for hydrocarbon exploration and development. The slope systems are characterized by seven seismic facies building blocks, including the turbiditic channel fills, the turbidite lobes, the sheet turbidites, the slide, slump and debris flow sheets, lobes and tongues, the fine-grained turbidite fills and sheets, the contourite drifts and finally, the hemipelagic drapes and fills. Sparker profiles offshore Ischia are presented. New seismo-stratigraphic evidence on buried volcanic structures and overlying Quaternary deposits of the eastern offshore of the Ischia Island are here discussed to highlight the implications on marine geophysics and volcanology. Regional seismic sections in the Ischia offshore across buried volcanic structures and debris avalanche and debris flow deposits are here presented and discussed. Deep sea depositional systems in the Ischia Island are well developed in correspondence to the Southern Ischia canyon system. The canyon system engraves a narrow continental shelf from Punta Imperatore to Punta San Pancrazio, being limited southwestwards from the relict volcanic edifice of the Ischia bank. While the eastern boundary of the canyon system is controlled by extensional tectonics, being limited from a NE-SW trending (counter-Apenninic) normal fault, its western boundary is controlled by volcanism, due to the growth of the Ischia volcanic bank. Submarine gravitational instabilities also acted in relationships to the canyon system, allowing for the individuation of large scale creeping at the sea bottom and hummocky deposits already interpreted as debris avalanche deposits. High resolution seismic data (Subbottom Chirp) coupled to high resolution Multibeam bathymetry collected in the frame of the Stromboli geophysical experiment aimed at recording seismic active data and tomography of the Stromboli Island are here presented. A new detailed swath bathymetry of Stromboli Island is here shown and discussed to reconstruct an up-to-date morpho-bathymetry and marine geology of the area, compared to volcanologic setting of the Aeolian volcanic complex. The Stromboli DEM gives information about the submerged structure of the volcano, particularly about the volcano-tectonic and gravitational processes involving the submarine flanks of the edifice. Several seismic units have been identified around the volcanic edifice and interpreted as volcanic acoustic basement pertaining to the volcano and overlying slide chaotic bodies emplaced during its complex volcano-tectonic evolution. They are related to the eruptive activity of Stromboli, mainly poliphasic and to regional geological processes involving the geology of the Aeolian Arc.

Cone penetration testing to assess slope stability in the 1979 Nice landslide area (Ligurian Margin, SE France)

In the landslide-prone area near the Nice international airport, southeastern France, an interdisciplinary approach is applied to develop realistic lithological/geometrical profiles and geotechnical/strength sub-seafloor models. Such models are indispensable for slope stability assessments using limit equilibrium or finite element methods. Regression analyses, based on the undrained shear strength (su) of intact gassy sediments are used to generate a sub-seafloor strength model based on 37 short dynamic and eight long static piezocone penetration tests, and laboratory experiments on one Calypso piston and 10 gravity cores. Significant strength variations were detected when comparing measurements from the shelf and the shelf break, with a significant drop in su to 5.5 kPa being interpreted as a weak zone at a depth between 6.5 and 8.5 m below seafloor (mbsf). Here, a 10% reduction of the in situ total unit weight compared to the surrounding sediments is found to coincide with coarse-grained layers that turn into a weak zone and detachment plane for former and present-day gravitational, retrogressive slide events, as seen in 2D chirp profiles. The combination of high-resolution chirp profiles and comprehensive geotechnical information allows us to compute enhanced 2D finite element slope stability analysis with undrained sediment response compared to previous 2D numerical and 3D limit equilibrium assessments. Those models suggest that significant portions (detachment planes at 20 m or even 55 mbsf) of the Quaternary delta and slope apron deposits may be mobilized. Given that factors of safety are equal or less than 1 when further considering the effect of free gas, a high risk for a landslide event of considerable size off Nice international airport is identified

The pianosa contourite depositional system (northern Tyrrhenian sea): drift morphology and plio-quaternary stratigraphic evolution

The Pianosa Contourite Depositional System (CDS) is located in the Corsica Trough (Northern Tyrrhenian Sea), a confined basin dominated by mass transport and contour currents in the eastern flank and by turbidity currents in the western flank. The morphologic and stratigraphic characterisation of the Pianosa CDS is based on multibeam bathymetry, seismic reflection data (multi-channel high resolution mini GI gun, single-channel sparker and CHIRP), sediment cores and ADCP data. The Pianosa CDS is located at shallow to intermediate water depths (170 to 850 m water depth) and is formed under the influence of the Levantine Intermediate Water (LIW). It is 120 km long, has a maximum width of 10 km and is composed of different types of muddy sediment drifts: plastered drift, separated mounded drift, sigmoid drift and multicrested drift. The reduced tectonic activity in the Corsica Trough since the early Pliocene permits to recover a sedimentary record of the contourite depositional system that is only influenced by climate fluctuations. Contourites started to develop in the Middle-Late Pliocene, but their growth was enhanced since the Middle Pleistocene Transition (0.7–0.9 Ma). Although the general circulation of the LIW, flowing northwards in the Corsica Trough, remained active all along the history of the system, contourite drift formation changed, controlled by sediment influx and bottom current velocity. During periods of sea level fall, fast bottom currents often eroded the drift crest in the middle and upper slope. At that time the proximity of the coast to the shelf edge favoured the formation of bioclastic sand deposits winnowed by bottom currents. Higher sediment accumulation of mud in the drifts occurred during periods of fast bottom currents and high sediment availability (i.e. high activity of turbidity currents), coincident with periods of sea level low-stands. Condensed sections were formed during sea level high-stands, when bottom currents were more sluggish and the turbidite system was disconnected, resulting in a lower sediment influx.

Electro-optical characterization of IC compatible microcantilevers

The aim of this work is to simulate and optically characterize the piezoelectric performance of complementary metal oxide semiconductor (CMOS) compatible microcantilevers based on aluminium nitride (AlN) and manufactured at room temperature. This study should facilitate the integration of piezoelectric micro-electro-mechanical systems (MEMS) such as microcantilevers, in CMOS technology. Besides compatibility with standard integrated circuit manufacturing procedures, low temperature processing also translates into higher throughput and, as a consequence, lower manufacturing costs. Thus, the use of the piezoelectric properties of AlN manufactured by reactive sputtering at room temperature is an important step towards the integration of this type of devices within future CMOS technology standards. To assess the reliability of our fabrication process, we have manufactured arrays of free-standing microcantilever beams of variable dimension and studied their piezoelectric performance. The characterization of the first out-of-plane modes of AlN-actuated piezoelectric microcantilevers has been carried out using two optical techniques: laser Doppler vibrometry (LDV) and white light interferometry (WLI). In order to actuate the cantilevers, a periodic chirp signal in certain frequency ranges was applied between the device electrodes. The nature of the different vibration modes detected has been studied and compared with that obtained by a finite element model based simulation (COMSOL Multiphysics), showing flexural as well as torsional modes. The correspondence between theoretical and experimental data is reasonably good, probing the viability of this high throughput and CMOS compatible fabrication process. To complete the study, X-ray diffraction as well as d33 piezoelectric coefficient measurements were also carried out.

Amplification d’impulsions brèves de haute énergie par effet Raman stimulé dans les fibres optiques

Le développement au cours des dernières décennies de lasers à fibre à verrouillage de modes permet aujourd’hui d’avoir accès à des sources fiables d’impulsions femtosecondes qui sont utilisées autant dans les laboratoires de recherche que pour des applications commerciales. Grâce à leur large bande passante ainsi qu’à leur excellente dissipation de chaleur, les fibres dopées avec des ions de terres rares ont permis l’amplification et la génération d’impulsions brèves de haute énergie avec une forte cadence. Cependant, les effets non linéaires causés par la faible taille du faisceau dans la fibre ainsi que la saturation de l’inversion de population du milieu compliquent l’utilisation d’amplificateurs fibrés pour l’obtention d’impulsions brèves dont l’énergie dépasse le millijoule. Diverses stratégies comme l’étirement des impulsions à des durées de l’ordre de la nanoseconde, l’utilisation de fibres à cristaux photoniques ayant un coeur plus large et l’amplification en parallèle ont permis de contourner ces limitations pour obtenir des impulsions de quelques millijoules ayant une durée inférieure à la picoseconde. Ce mémoire de maîtrise présente une nouvelle approche pour l’amplification d’impulsions brèves utilisant la diffusion Raman des verres de silice comme milieu de gain. Il est connu que cet effet non linéaire permet l’amplification avec une large bande passante et ce dernier est d’ailleurs couramment utilisé aujourd’hui dans les réseaux de télécommunications par fibre optique. Puisque l’adaptation des schémas d’amplification Raman existants aux impulsions brèves de haute énergie n’est pas directe, on propose plutôt un schéma consistant à transférer l’énergie d’une impulsion pompe quasi monochromatique à une impulsion signal brève étirée avec une dérive en fréquence. Afin d’évaluer le potentiel du gain Raman pour l’amplification d’impulsions brèves, ce mémoire présente un modèle analytique permettant de prédire les caractéristiques de l’impulsion amplifiée selon celles de la pompe et le milieu dans lequel elles se propagent. On trouve alors que la bande passante élevée du gain Raman des verres de silice ainsi que sa saturation inhomogène permettent l’amplification d’impulsions signal à une énergie comparable à celle de la pompe tout en conservant une largeur spectrale élevée supportant la compression à des durées très brèves. Quelques variantes du schéma d’amplification sont proposées, et leur potentiel est évalué par l’utilisation du modèle analytique ou de simulations numériques. On prédit analytiquement et numériquement l’amplification Raman d’impulsions à des énergies de quelques millijoules, dont la durée est inférieure à 150 fs et dont la puissance crête avoisine 20 GW.