RCAAP Repository

This paper presents a baseline system for automatic acoustic scene classification based on the audio signals alone. The proposed method  is derived from classic, content-based, music classification approaches, and consists in a feature extraction phase followed by two dimensionality reduction steps (PCA and LDA) and a classification phase done using a k nearest-neighbors algorithm.This paper also reports on how our system performed  in the context of the DCASE 2016 challenge, for the acoustic scene classification task.

This paper address the implementation of a mixed-domain simulator for first-order band-pass sampling receivers, which is based on an initial frequency-domain signal treatment followed by a time-domain simulation scheme. One of the proposed applications for this type of receivers is to perform the spectrum sensing feature, which is required in actual and future cognitive radio approaches. Some details about the multi-stage modelling strategy will be given focusing in each specific component of the receiver, wherein it is considered a mixed frequency-time signal treatment. Moreover, it will be summarized the main features of the implemented simulator, as well as potential improvements. Finally, several simulation examples obtained with the implemented simulator will be shown, in which are included the impact of a CW signal excitation in a received modulated signal and a multi-carrier signal reception scenario.

The search for materials for hydrogen storage that are able to adsorb and desorb hydrogen molecules has received attention. Nanofibers have been studied because it has some interesting characteristics like high surface to volume ratio and the ease of producing by means of a low-cost setup. Polyacrylonitrile (PAN) has allowed the possibility of obtaining carbon nanofibers from the electrospinning process followed by a thermal treatment; moreover Pd has the capability of store hydrogen and can form a complex with dimethylformamide (DMF) so that, during the preparation of the solution of PAN and PdCl2, both present high solubility in the solvent.  Thus, the results demonstrated that as Pd can form complex with DMF the preparation method of the PAN+PdCl2 solutions influences the interaction between Pd and PAN. The decrease in dispersion viscosity and an unexpected increase in the diameter of the fibers confirm the hypothesis. The FTIRS and XPS analysis confirm that the Pd complexes with DMF by means of bonds to N.

In text classification based on the bag-of-words (BoW) or similar representations, we usually have a large number of features, many of which are irrelevant (or even detrimental) for classification tasks. Recent results show that compressed learning (CL), i.e., learning in a domain of reduced dimensionality obtained by random projections (RP), is possible, and theoretical bounds on the test set error rate have been shown. In this work, we assess the performance of CL, based on RP of BoW representations for text classification. Our experimental results show that CL significantly reduces the number of features and the training time, while simultaneously improving the classification accuracy. Rather than the mild decrease in accuracy upper bounded by the theory, we actually find an increase of accuracy. Our approach is further compared against two techniques, namely the unsupervised random subspaces method and the supervised Fisher index. The CL approach is suited for unsupervised or semi-supervised learning, without any modification, since it does not use the class labels.

Physical devices with different sensors and sampling rates distributed over several unrelated locations need to store their values over time. Applications that need the result of a set of sensors must access their data. A common and simple interface within the physical devices, monitoring stations, to store data on a database is needed; as also a simple and common retrieval interface for any application that only shows the data as it is or processes it into higher levels of significance. BlueLab IoT is a platform with libraries and an interface application to aid that development; a working implementation is provided.

This article reports on a monolithic 10 cm × 10 cm area PV module integrating an array of 72 a-Si:H n-i-p cells on a thin polyethylene-naphtalate substrate. The design optimization and device performance analysis are performed using a two-dimensional distributed circuit model of the photovoltaic cell. Experimental results show that the shunt leakage is one of the factors reducing the device performance. Using the LBIC technique, the multiple micro-shunts in the n‑i‑p cell were detected. The mechanism of electrical shunts formation is proposed and discussed.

In this work, an in-depth analysis concerning the transmission performance of IEEE802.11g/n (Wi-Fi) signals in a WDM-PON system is presented. It is considered that the optical/electrical transceivers are based on low-cost 850 nm VCSELs and PIN photodiodes. System modelling includes the impact of noise generated in the optical path, such as relative intensity noise (RIN), shot noise, photodetector thermal noise, clipping and intermodulation distortion. An analytic analysis based on Volterra series is conducted and mathematical expressions for both the EVM and SNR are derived. The theoretical analysis is also compared with experimental results. Among several conclusions, it is observed that the laser intermodulation distortion, clipping and RIN are the most relevant factors.

This paper presents the manufacturing of amicroelectrode array, on printed circuit boards (PCB) and onsilicon substrate, composed of 60 to 64 gold microelectrodes(between 4 µm and 70 µm each in diameter). The sensor has ameasurement total area of 0.5 cm in radius, one referenceelectrode (1 mm2 in area) and 28.4 mm wide and 28.4 mm long.These microelectrodes were used for checking and logging ofextracelullar local field potential of cell culture and nitritemeasurement in a phosphate buffered saline solution(electrolytical aqueous medium). In addition, an apparatus toshield from electromagnetic interference for connecting thearrays was designed to allow the capture of electrochemicalreactions or electronic signals by the microelectrodes, forexample: nitrite or cardiac potential measurement, respectively.Finally, biocompatibility tests of the array structures wereperformed. The preliminary electrical and biocompatibilitytesting, along with the collected data, has shown promisingresults pointing to the development of an accurate sensor afterthe completion of this study. The sensor has potentially abroader range of applications with only a few adaptations anddue its good accuracy it can be a very useful resource for manychemical and biological applications.

This paper analyses the Alamouti scheme for different antenna configurations and different modulation types,namely BPSK, QPSK and QAM. All configurations were modeled and simulated in MATLAB. A MIMOreceiver for a 21 antenna configuration and BPSK modulation was implemented in a FPGA. The FPGAresults indicate that the Alamouti scheme is a good design option for hardware implementation of a MIMOreceiver. The receiver uses only about 10% of the resources of a medium-sized FPGA and achieves almost300 Msymbols per second.

The emerging concept of smart cities demands for a large number of electronic devices, like sensors and actuators, distributed over several public spaces and buildings. The Internet of Things (IoT) has a key role in connecting devices to the Internet. However, the significant number of devices makes the maintenance task of the entire network difficult and expensive. To mitigate this problem, considerable research efforts have been made to develop energy-aware devices capable of self-sustainable operation, by harvesting their energy from various sources. In this paper, we study the possibility of harvesting energy from the light flowing in the Gigabit Passive Optics Network (GPON) to supply low-power devices. Since most cities already have a working GPON installation, using this installation to interconnect and power IoT devices can be a viable and less expensive solution, instead of installing new dedicated networks. This is also an interesting solution to convey communications and energy to low-power applications where access to the power grid is unfeasible. This study is focused in the 1550 nm wavelength, whose available optical power, in residential premises, is between -7 dBm and +2 dBm. With this range of optical power, and with a 30% efficiency photodiode, we show, for the worst-case scenario of the GPON, how it is possible to harvest 62 µW of energy at the Maximum Power Point (MPP).

This paper presents a new single cell multi-objectiveoptimization algorithm. The objective is to optimize areas oflow coverage and high interference simultaneously, through theadjustment of the antenna tilts and/or antenna orientation. Theprocess is achieved using a specific implementation of a ParticleSwarm Optimization (PSO) algorithm. Both the detection ofsub-optimal performance areas and its subsequent optimizationare supported by Drive Test (DT) data and network topologyinformation. The antenna optimization algorithm was tested withreal data in 3rd Generation (3G)/4th Generation (4G) networks.In this work, a 3G urban scenario is highlighted, achieving anaverage optimization gain of 78%.

This work presents fully automated plasma-enhanced reactive thermal evaporation system (rf-PERTE) that can be used for the deposition of transparent metal oxide films with high reproducibility of their electrical and optical properties. The developed hardware/software platform enables the full control over the critical deposition conditions such as mass flow of oxygen, process pressure, current flowing through crucible and rf-power. For indium oxide films on glass substrates a resistivity of 9×10-4 Ω-cm and a transmittance of 90% in the visible spectral range were achieved without substrate heating. The system is also suitable for the deposition of transparent conducting coatings in a wide range of plastic substrates, for applications in the field of flexible sensors or solar cells. In particular, we have successfully deposited indium oxide on PEN (polyethylene naphthalate) sheets with electrical and optical properties approaching the ones of the films deposited on glass substrates.

In this paper we report the use of a monolithic system that combines the demultiplexing operation with the simultaneous photodetection and self amplification of the signal. The device is a double pi’n/pin a-SiC:H heterostructure with optical gate connections for light triggering in different spectral regions. Results show that when a polychromatic combination of different pulsed channels impinges on the device the output signal has a strong nonlinear dependence on the light absorption profile, (wavelength, bit rate and intensity). This effect is due to the self biasing of the junctions under unbalanced light generation of carriers. Self optical bias amplification under uniform irradiation and transient conditions is achieved. An optoelectronic model based on four essential elements: a voltage supply, a monolithic double pin photodiode, optical connections for light triggering, and optical power sources for light bias explains the operation of the optical system.

Web Archiving (WA) deals with the preservation of portions of the World Wide Web (WWW) allowing their availability for future access. Arquivo.pt is a WA initiativeholding a huge amount of content, including image files.However, some of these images contain nudity and pornography, that can be offensive for the users, and thus being Not SuitableFor Work (NSFW). This work proposes a methodology to classify NSFW images available at Arquivo.pt, using deep neural network approaches. A large dataset of images is built using Arquivo.pt data and two pre-trained neural network models, namely ResNet and SqueezeNet, are evaluated and improved for the NSFW classification task, using the dataset.The evaluation of these models reported an accuracy of 93% and 72%, respectively. After a fine tuning stage, the accuracy of these models improved to 94% and 89%, respectively.The proposed solution is integrated into the Arquivo.pt Image Search System, enabling the filtering of the problematic NSFW images. At the time of this writing, the proposed solution is in production at https://arquivo.pt/images.jsp

The transducer consists on a semiconductor device based on two stacked-i-n  heterostructures, that were designed to detect the emissions of the fluorescence resonance energy transfer between fluorophores in the cyan (470 nm) and yellow (588 nm) range of the spectrum. This research represents a preliminary study on the use of such wavelength sensitive devices as photodetectors for this kind of application. The device was characterized through optoelectronic measurements concerning spectral response measurements under different electrical and optical biasing conditions. To simulate the FRET pairs a chromatic time dependent combination of cyan and yellow wavelengths was applied to the device. The generated photocurrent was measured under reverse and forward bias to readout the output photocurrent signal. Different wavelength biasing light was also superimposed. Results show that under reverse bias the photocurrent signal presents four separate levels each one assigned to the different wavelength combinations of the FRET pairs. If a blue background is superimposed the yellow channel is enhanced and the cyan suppressed while under red irradiation the opposite behavior occurs. So under suitable biasing light the transducer is able to detect separately the cyan and yellow fluorescence pairs. An electrical model, supported by a numerical simulation supports the transduction mechanism of the device.

a use case of poultry counting on a production line in Portugal is presented as a first step for Industry 4.0 application. Along the poultry production line there are several transformation and translation steps in which the fowls are subjected to, and their quantity in the line may change; some are removed or fall out and others are recovered and inserted in precise places of the line. Those changes can be used as an indicator of the status of the production line at that may influence the improvement of its efficiency. This paper presents a working solution of five wireless devices along a poultry production line and the complete system where data is stored locally for maintenance and supervisor decisions, and also data is stored at a remote database for management purposes which can be accessed by a mobile application.

 Abstract1— Optical coherence tomography (OCT) is a non-invasive biomedical imaging technique that provides high speed and high resolution three dimensional and cross sectional images of biological samples, in vivo and in situ. OCT applications targeting small animals is believed to bring developments in medical techniques, instruments, diagnosis and therapies for a number of human diseases as always have been the case of animal experimentation. With the swept source OCT (SS-OCT) system presented in this work, we were able to achieve performance parameters that meet the requirements to image the retina of small animals. Performance characteristics include 105 dB for system sensitivity, a roll-off below 1 dB/mm over 3 mm depth and an axial resolution of 8 μm. We describe the layout and acquisition/processing solutions towards fast imaging of in vivo samples.

This paper presents a study on resonant converter topologies targeted for CMOS integration. Design methodologies to optimize efficiency for the integration of Quasi-Resonant and Quasi-Square-Wave converters are proposed. A power loss model is used to optimize the design parameters of the power stage, including the driver circuits, and also to conclude about CMOS technology limitations. Based on this discussion, and taking as reference a 0.35μm CMOS process, two converters are designed to validate the proposal: a Quasi Resonant boost converter operating at 100MHz and a Quasi-Square-Wave buck converter operating at 70MHz. Simulation results confirm the feasibility of these topologies for monolithic integration.

The interest in monitoring a drivers conditions and performance has increased in the past years, to make the roads safer both for drivers and pedestrians. This raised the idea of developing a system to monitor the drivers conditions to prevent road disasters. In this paper, we propose a system to monitor the drivers fatigue and drowsiness, based on the Car- dioWheel system, developed by CardioID. The proposed system records both the persons ECG signal and the motion of the steering wheel during the driving session. The amount of data acquired demands a compression stage for transmission with the goal to reduce the required bandwidth. The transmission of the compressed data is done via Bluetooth Low Energy, with an exclusive profile developed for this system. To detect fatigue and drowsiness patterns, a machine learning approach was taken. Among the evaluated classifiers, the Support Vector Machines technique proved to be the best classification method with the highest accuracy. Thus, the developed prototype has the ability to warn the driver about his physiological and physical states, increasing the safety in the roads.

With the first 3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE) networks being deployed more complexity is added to current existing cellular mobile networks and more capital (CAPEX) and operational (OPEX) effort will be needed. In addition, the rising demand of users for new services and higher data rates demands more efficiency from operators. For this matter, 3GPP Release 8 as introduced the Self-Organizing Network (SON) concept, a set of self-configuration, self-optimizing and self-healing functions that allow the automation of labor-intensive tasks, reducing operational and capital costs. While requirements on cutting operational expenditure remain, operators still remain skeptical with the efficiency of these functions. In this paper, Physical Cell Identity (PCI) conflict detection and resolution, Automatic Neighbor Relation (ANR) and automatic Handover Parameter Optimization (HPO) functions are proposed as part of a simulator for LTE SON based networks. Based on user defined inputs, these functions allow operators to closely predict and gather optimal policy input values for SON algorithms, while maintaining desirable network performance. Based on a real network scenario, results show simulator’s clear benefit when compared with other proposals.