3D image reconstruction with electron tomography holds problems due to the severely limited range of projection angles and low signal to noise ratio of the acquired projection images. The maximum a posteriori (MAP) reconstruction methods have been successful in compensating for the missing information and suppressing noise with their intrinsic regularization techniques. There are two major problems in MAP reconstruction methods: (1) selection of the regularization parameter that controls the balance between the data fidelity and the prior information, and (2) long computation time. One aim of this study is to provide an adaptive solution to the regularization parameter selection problem without having additional knowledge about the imaging environment and the sample. The other aim is to realize the reconstruction using sequences of resolution levels to shorten the computation time. The reconstructions were analyzed in terms of accuracy and computational efficiency using a simulated biological phantom and publically available experimental datasets of electron tomography. The numerical and visual evaluations of the experiments show that the adaptive multiresolution method can provide more accurate results than the weighted back projection (WBP), simultaneous iterative reconstruction technique (SIRT), and sequential MAP expectation maximization (sMAPEM) method. The method is superior to sMAPEM also in terms of computation time and usability since it can reconstruct 3D images significantly faster without requiring any parameter to be set by the user.

In this paper, we propose two novel methods for robot-world-hand-eye calibration and provide a comparative analysis against six state-of-the-art methods. We examine the calibration problem from two alternative geometrical interpretations, called 'hand-eye' and 'robot-world-hand-eye', respectively. The study analyses the effects of specifying the objective function as pose error or reprojection error minimization problem. We provide three real and three simulated datasets with rendered images as part of the study. In addition, we propose a robotic arm error modeling approach to be used along with the simulated datasets for generating a realistic response. The tests on simulated data are performed in both ideal cases and with pseudo-realistic robotic arm pose and visual noise. Our methods show significant improvement and robustness on many metrics in various scenarios compared to state-of-the-art methods.

Respiration is vital to land-dwelling mammals; besides, salient information is encoded in the respiratory rate. Objective assessment of the respiratory rate is difficult in dogs; in particular, if the unobtrusive measurement is desired. The goal of this work was to develop and evaluate a method for unobtrusive sensing of respiratory rate in dogs. For this, the 'FlexPock' multisensor system, originally developed for unobtrusive estimation of heart rate and respiratory rate in humans via magnetic impedance; accelerometry; and optical measurements, was used to assess canine respiratory rate. In a proof-of-concept study with 10 healthy dogs of different breeds and sizes, a total of 240 minutes of data was recorded in the phases standing, sitting, lying down, and walking. An algorithm was developed that estimates the respiratory rate by fusing the information from multiple sensors for increased accuracy and robustness. To discard unusable data, a simple yet effective signal quality metric was introduced. Impedance pneumography recorded using adhesive electrodes was used as a reference. Analysis of the raw FlexPock data revealed that the magnetic impedance and accelerometry were the best individual sensing modalities and fusion of these data further increased the accuracy. Using leave-one-dog-out cross-validation, the average estimation error was 9.5% at a coverage of 50.1%. However, strong variation between dogs and phases was observed. During the walking phase, neither reference nor unobtrusive sensor reported usable results, while the sitting phase exhibited the best performance. In conclusion, the fusion of magnetic impedance and accelerometry can be used for unobtrusive respiratory rate estimation in stationary dogs.

Here, we describe the use of a biolayer interferometry biosensor for the fast and sensitive detection of virus-specific antibodies from human serum samples. Norovirus-like particles and norovirus P-particles were used to functionalise the biosensor tip. The detection of antibodies directly from serum samples was challenging, but the addition of a metal chelator (DAB) combined with an anti-human horseradish peroxidase-tagged antibody enabled enhanced detection of virus-specific antibodies in serum dilutions up to 1:100,000. Biolayer interferometry provides results faster than an ELISA, with results in as little as 10-20 min when using pre-functionalised sensors. Therefore, biolayer interferometry combined with DAB enhancement offers an attractive method for quick and sensitive quantification of biomolecules from complicated sample matrices.

Exemplar-based speech enhancement systems work by decomposing the noisy speech as a weighted sum of speech and noise exemplars stored in a dictionary and use the resulting speech and noise estimates to obtain a time-varying filter in the full-resolution frequency domain to enhance the noisy speech. To obtain the decomposition, exemplars sampled in lower dimensional spaces are preferred over the full-resolution frequency domain for their reduced computational complexity and the ability to better generalize to unseen cases. But the resulting filter may be sub-optimal as the mapping of the obtained speech and noise estimates to the full-resolution frequency domain yields a low-rank approximation. This paper proposes an efficient way to directly compute the full-resolution frequency estimates of speech and noise using coupled dictionaries: an input dictionary containing atoms from the desired exemplar space to obtain the decomposition and a coupled output dictionary containing exemplars from the full-resolution frequency domain. We also introduce modulation spectrogram features for the exemplar-based tasks using this approach. The proposed system was evaluated for various choices of input exemplars and yielded improved speech enhancement performances on the AURORA-2 and AURORA-4 databases. We further show that the proposed approach also results in improved word error rates (WERs) for the speech recognition tasks using HMM-GMM and deep-neural network (DNN) based systems.

This paper presents an algorithm for unsupervised single-channel source separation of audio mixtures. The approach specifically addresses the challenging case of separation where no training data are available. By representing mixtures in the modulation spectrogram (MS) domain, we exploit underlying similarities in patterns present across frequency. A three-dimensional tensor factorization is able to take advantage of these redundant patterns, and is used to separate a mixture into an approximated sum of components by minimizing a divergence cost. Furthermore, we show that the basic tensor factorization can be extended with convolution in time being used to improve separation results and provide update rules to learn components in such a manner. Following factorization, sources are reconstructed in the audio domain from estimated components using a novel approach based on reconstruction masks that are learned using MS activations, and then applied to a mixture spectrogram. We demonstrate that the proposed method produces superior separation performance to a spectrally based nonnegative matrix factorization approach, in terms of source-to-distortion ratio. We also compare separation with the perceptually motivated interference-related perceptual score metric and identify cases with higher performance.

In breast cancer screening, the radiation dose must be kept to the minimum necessary to achieve the desired diagnostic objective, thus minimizing risks associated with cancer induction. However, decreasing the radiation dose also degrades the image quality. In this work we restore digital breast tomosynthesis (DBT) projections acquired at low radiation doses with the goal of achieving a quality comparable to that obtained from current standard full-dose imaging protocols. A multiframe denoising algorithm was applied to low-dose projections, which are filtered jointly. Furthermore, a weighted average was used to inject a varying portion of the noisy signal back into the denoised one, in order to attain a signal-to-noise ratio comparable to that of standard full-dose projections. The entire restoration framework leverages a signal-dependent noise model with quantum gain which varies both upon the projection angle and on the pixel position. A clinical DBT system and a 3D anthropomorphic breast phantom were used to validate the proposed method, both on DBT projections and slices from the 3D reconstructed volume. The framework is shown to attain the standard full-dose image quality from data acquired at 50% lower radiation dose, whereas progressive loss of relevant details compromises the image quality if the dosage is further decreased.

The prism coupling technique has been utilized to measure the refractive index in the near- and mid-IR spectral region of chalcogenide glasses in bulk and thin film form. A commercial system (Metricon model 2010) has been modified with additional laser sources, detectors, and a new GaP prism to allow the measurement of refractive index dispersion over the 1.5-10.6m range. The instrumental error was found to be 0.001 refractive index units across the entire wavelength region examined. Measurements on thermally evaporated AMTIR2 thin films confirmed that (i) the film deposition process provides thin films with reduced index compared to that of the bulk glass used as a target, (ii) annealing of the films increases the refractive index of the film to the level of the bulk glass used as a target to create it, and (iii) it is possible to locally increase the refractive index of the chalcogenide glass using laser exposure at 632.8 nm.

In this paper, we establish a new type of passive ultra-high-frequency (UHF) radio frequency identification (RFID) technology-based moisture sensor. The sensor system consists of a sensor tag and a reference tag to indicate exposure to moisture. The sensor tag is fabricated on a cotton substrate from conductive silver yarn and water-soluble polyvinyl alcohol (PVA) yarn. When the sensor is exposed to water, the PVA yarn will melt, which causes a permanent change to the tag antenna body and to the wireless performance of the sensor tag. By measuring the backscattered power strength difference, ΔP, the presence of moisture can be detected and recorded. When ΔP is below 0, the sensor system shows that it has been in a highly moist condition. Our PVA-based moisture-sensing system can provide cost-effective and maintenance-free monitoring of moisture exposure for versatile application fields.

Radio frequency identification (RFID) technology has enabled a new class of low cost, wireless zero-power sensors, which open up applications in highly pervasive and distributed RFID-enabled sensing, which were previously not feasible with wired or battery powered wireless sensor nodes. This paper provides a review of RFID sensing techniques utilizing chip-based and chipless RFID principles, and presents a variety of implementations of RFID-based sensors, which can be used to detect strain, temperature, water quality, touch, and gas.

In this study, we propose an unsupervised method for dictionary learning in audio signals. The new method, called binary nonnegative matrix deconvolution (BNMD), is developed and used to discover patterns from magnitude-scale spectrograms. The BNMD models an audio spectrogram as a sum of delayed patterns having binary gains (activations). Only small subsets of patterns can be active for a given spectrogram excerpt. The proposed method was applied to speaker identification and separation tasks. The experimental results show that dictionaries obtained by the BNMD bring much higher speaker identification accuracies averaged over a range of SNRs from -6 dB to 9 dB (91.3%) than the NMD-based dictionaries (37.8-75.4%). The BNMD also gives a benefit over dictionaries obtained using vector quantization (87.8%). For bigger dictionaries the difference between the BNMD and the vector quantization (VQ) is getting smaller. For the speech separation task the BNMD dictionary gave a slight improvement over the VQ.

The wind field is an essential factor that affects accurate homing and flare landing of parafoil systems. In order to obtain the ambient wind field during the descent of a parafoil system, a combination method of in-flight wind field identification and prediction is proposed. First, a wind identification method only using global position system information is derived based on the flight dynamics of parafoil systems. Then a wind field prediction model is constructed using the atmospheric dynamics, and the low-altitude wind field is predicted based on the identified wind field of high-altitude. Finally, simulations of wind field identification and prediction are conducted. The results demonstrate that the proposed method can identify the wind fields precisely and also predict the wind fields reasonably. This method can potentially be applied in practical parafoil systems to provide wind field information for homing tasks.

This paper presents the design of low-cost, conformal UHF antennas and RFID tags on two types of cork substrates: 1) natural cork and 2) agglomerate cork. Such RFID tags find an application in wine bottle and barrel identification, and in addition, they are suitable for numerous antenna-based sensing applications. This paper includes the high-frequency characterization of the selected cork substrates considering the anisotropic behavior of such materials. In addition, the variation of their permittivity values as a function of the humidity is also verified. As a proof-of-concept demonstration, three conformal RFID tags have been implemented on cork, and their performance has been evaluated using both a commercial Alien ALR8800 reader and an in-house measurement setup. The reading of all tags has been checked, and a satisfactory performance has been verified, with reading ranges spanning from 0.3 to 6 m. In addition, this paper discusses how inkjet printing can be applied to cork surfaces, and an RFID tag printed on cork is used as a humidity sensor. Its performance is tested under different humidity conditions, and a good range in excess of 3 m has been achieved, allied to a good sensitivity obtained with a shift of >5 dB in threshold power of the tag for different humid conditions.

This article demonstrates that the number of pulses in a group of vector bound solitons can be accurately adjusted by polarization control. We could achieve states with up to seven pulses trapped in a group with equally firm pulse separation. The stable relative phase between pulses within the group is manifested as an interference pattern in the soliton spectrum. © 2013 Astro Ltd.

Electrically tunable metasurfaces with graphene offer design flexibility to efficiently manipulate and control light. These metasurfaces can be used to generate plasmon-induced reflectance (PIR), which can be tuned by electrostatic doping of the graphene layer. We numerically investigated two designs for tunable PIR devices using the finite difference time-domain (FDTD) method. The first design is based on two rectangular antennas of the same size and a disk; in the second design, two parallel rectangular antennas with different dimensions are used. The PIR-effect was achieved by weak hybridization of two bright modes in both devices and tuned by changing the Fermi level of graphene. A total shift of 362 nm was observed in the design with the modulation depth of 53% and a spectral contrast ratio of 76%. These tunable PIR devices can be used for tunable enhanced biosensing and switchable systems.

This paper introduces a prototype of ClothFace technology, a battery-free textile-based handwriting recognition platform that includes an e-textile antenna and a 10 × 10 array of radio frequency identification (RFID) integrated circuits (ICs), each with a unique ID. Touching the textile platform surface creates an electrical connection from specific ICs to the antenna, which enables the connected ICs to be read with an external UHF (ultra-haigh frequency) RFID reader. In this paper, the platform is demonstrated to recognize handwritten numbers 0-9. The raw data collected by the platform are a sequence of IDs from the touched ICs. The system converts the data into bitmaps and their details are increased by interpolating between neighboring samples using the sequential information of IDs. These images of digits written on the platform can be classified, with enough accuracy for practical use, by deep learning. The recognition system was trained and tested with samples from six volunteers using the platform. The real-time number recognition ability of the ClothFace technology is demonstrated to work successfully with a very low error rate. The overall recognition accuracy of the platform is 94.6% and the accuracy for each digit is between 91.1% and 98.3%. As the solution is fully passive and gets all the needed energy from the external RFID reader, it enables a maintenance-free and cost-effective user interface that can be integrated into clothing and into textiles around us.

Comparative study of the nonlinear optical response of pyrrolo- tetrathiafulvalene polygons has been carried out. These materials are of great importance for photonics, because they can be used as building blocks for more complicated molecular systems like rotaxanes, catenanes which are promising candidates for opto-electronic applications like optical computing, optical switching etc. During this work the third order nonlinear susceptibility (X ⁽³⁾) has been determined by the Degenerate Four Wave Mixing experiment, while the imaginary part of the X ⁽³⁾, which is related with the nonlinear absorption was determined by "open-aperture" Z-scan measurements. Moreover measurements have been carried out in order to provide information about the optical limiting efficiency of the materials. High X ⁽³⁾ has been found reaching 2.30 × 10 ^-21 m ²/V ², corresponding to second order hyperpolarizability γ values of about 6.5 × 10 ^-45 m ⁵/V ².

Chronic wounds impose a significant financial burden for the healthcare system. Currently, assessment and monitoring of hard-to-heal wounds are often based on visual means and measuring the size of the wound. The primary wound dressings must be removed before assessment can be done. We have developed a quasi-monopolar bioimpedance-measurement-based method and a measurement system to determine the status of wound healing. The objective of this study was to demonstrate that with an appropriate setup, long-term monitoring of wound healing from beneath the primary dressings is feasible. The developed multielectrode sensor array was applied on the wound area and left under the primary dressings for 142 h. The impedance of the wounds and the surrounding intact skin area was measured regularly during the study at 150 Hz, 300 Hz, 1 kHz, and 5 kHz frequencies. At the end of the follow-up period, the wound impedance had reached the impedance of the intact skin at the higher frequencies and increased significantly at the lowest frequencies. The measurement frequency affected the measurement sensitivity in wound monitoring. The skin impedance remained stable over the measurement period. The sensor array also enabled the administration of periodical low-intensity direct current (LIDC) stimulation in order to create an antimicrobial environment across the wound area via the controlled formation of hydrogen peroxide (H2O2).

When exposed to air, alpha particles cause the production of light by exciting the molecules surrounding them. This light, the radioluminescence, is indicative of the presence of alpha radiation, thus allowing for the optical sensing of alpha radiation from distances larger than the few centimeters an alpha particle can travel in air. While the mechanics of radioluminescence in air and other gas compositions is relatively well understood, the same cannot be said about the radioluminescence properties of liquids. Better understanding of the radioluminescence properties of liquids is essential to design methods for the detection of radioactively contaminated liquids by optical means. In this article, we provide radioluminescence images of Am-241 dissolved in aqueous nitric acid (HNO ₃ ) solution and present the recorded radioluminescence spectrum with a maximum between and, and a steep decrease at the short wavelength side of the maximum. The shape of the spectrum resembles a luminescence process rather than Cerenkov light, bremsstrahlung, or other mechanisms with broadband emission. We show that the amount of light produced is about 150 times smaller compared to that of the same amount of Am-241 in air. The light production in the liquid is evenly distributed throughout the sample volume with a slight increase on the surface of the liquid. The radioluminescence intensity is shown to scale linearly with the Am-241 concentration and not be affected by the HNO ₃ concentration.

A fully passive, compact, and low-cost capacitive wireless radio frequency identification (RFID)-enabled sensing system for capacitive sensing and other Internet of Things applications is proposed. This calibration-free sensor utilizes a dual-tag topology, which consists of two closely spaced RFID tags with dipole antennas and printed capacitive sensor component connected to one of the tags. A series LC resonator is used to both reduce the antenna size and improve the isolation between the two antennas and the design/optimization steps are discussed in detail. All components except for the RFID chips are inkjet printed on an off-the-shelf photopaper using a silver nanoparticle ink. The complete sensor dimension is 84mm × 95mm and the sensor is compatible with EPC Class 1 Gen 2 (UHF) standard reader technology at 915 MHz.

Castable elastomers have been used to fabricate soft robotic devices and it has been shown that the technique scales well from prototyping to mass manufacturing. However, similarly scalable techniques for integrating strain or curvature sensors into such devices are still lacking. In this paper, we show that screenprinted silver conductors serve well as curvature sensors for soft robotic devices. The sensors are produced onto elastomer substrates in a single printing step and integrated into soft pneumatic actuators. We characterized the resistance-curvature relationship of the sensors, which allows the curvature of the actuators to be estimated from the sensor measurements. Hysteresis was observed, which does limit the absolute accuracy of the sensors. However, temperature characterizations showed that the sensor measurements are not significantly affected by temperature fluctuations during normal operation. Dynamic experiments showed that the bandwidth of the sensors is larger than the bandwidth of the actuators. We experimentally validated that these sensors can be used to detect whether the motion of an actuator has been blocked, clearing the way towards simple-tofabricate soft robots that react to their surroundings. Finally, we demonstrate a three-fingered soft robotic gripper with integrated sensors. We conclude that screen-printing is a promising way to integrate curvature sensors into soft robots.

This paper proposes, for the first time without using any linearization or order reduction, an adaptive and model-based discharge pressure control design for the variable displacement axial piston pumps (VDAPPs), whose dynamical behaviors are highly nonlinear and can be described by a fourth-order differential equation. The rigorous stability proof, with an asymptotic convergence, is given for the entire system. In the proposed novel controller design method, the specifically designed stabilizing terms constitute an essential core to cancel out all the stability-preventing terms. The experimental results reveal that rapid parameter adaptation significantly improves the feedback signal tracking precision compared to a known-parameter controller design. In the comparative experiments, the adaptive controller design demonstrates the state-of-the-art discharge pressure control performance, enabling a possibility for energy consumption reductions in hydraulic systems driven with VDAPP.

Since spring 2019 an experimental setup consisting of an electron spectrometer and an ion time-of-flight mass spectrometer for diluted samples has been available for users at the FinEstBeAMS beamline of the MAX IV Laboratory in Lund, Sweden. The setup enables users to study the interaction of atoms, molecules, (molecular) microclusters and nanoparticles with short-wavelength (vacuum ultraviolet and X-ray) synchrotron radiation and to follow the electron and nuclear dynamics induced by this interaction. Test measurements of N2 and thiophene (C4H4S) molecules have demonstrated that the setup can be used for many-particle coincidence spectroscopy. The measurements of the Ar 3p photoelectron spectra by linear horizontal and vertical polarization show that angle-resolved experiments can also be performed. The possibility to compare the electron spectroscopic results of diluted samples with solid targets in the case of Co2O3 and Fe2O3 at the Co and Fe L 2,3-absorption edges in the same experimental session is also demonstrated. Because the photon energy range of the FinEstBeAMS beamline extends from 4.4 eV up to 1000 eV, electron, ion and coincidence spectroscopy studies can be executed in a very broad photon energy range.

We propose an advanced scheme of amplifier similariton laser providing an output pulse spectrum much wider than the gain bandwidth. The upgrade is an additional dispersive element introduced into the cavity to locally increase the peak pulse power. The proposed scheme demonstrates a drastic increase in the output pulse spectrum width, reduction of the pulse duration, and an increase in the output peak pulse power after compression.

A variational method is used to derive numerical models for transient flow simulation in fluid transmission lines. These are generalizations of models derived using the more traditional modal method. Three different transient compressible laminar pipe flow models are considered (inviscous, one-dimensional linear viscous, and two-dimensional dissipative viscous flow), and a model for transient turbulent pipe flow is given. The (model) equations in the laminar case are given in the form of a set of constant coefficient ordinary differential equations, and for the turbulent case (model) in the form of a set of nonlinear ordinary differential equations. Explicit equations are given for various end conditions. Attenuation factors, similar to the window functions used in spectral analysis, are used to attenuate Gibbs phenomenon oscillations.

In this paper, a walk-through metal detection (WTMD) portal is used for classification of metallic objects. The classification is based on the inversion of the magnetic polarisability tensor (tensor) of the object. The nature of bias and noise components in the tensor are examined by using real walk-through data, and consequently, a novel classifier is introduced. Furthermore, a novel method for detecting poorly inverted tensors is presented, enabling self-diagnostics for the WTMD portal. Based on the results, the novel methods increase the accuracy of metal object classification and have the potential to improve the reliability of a WTMD system.

In this paper, a review of recent improvements on inkjet-printed microfluidic-based tunable/sensing RF systems is reported. The devices, such as Radio Frequency IDentification (RFID) passive wireless tags, coplanar patch antennas, bandstop filters, and loop antennas, are all fabricated by combining the inkjet printing technology on photographic paper for metallization and bonding layers, and laser etching for cavities and channels manufacturing. A novelty is also introduced for the loop antennas where the photographic paper is replaced with a polymer based substrate [i.e., (Poly(methyl-methacrylate))], to reduce the substrate losses for the RF part and solve the issue of paper hydrophylia. Along this paper an evolution toward higher working frequencies and higher detecting performance is shown, demonstrating a sensitivity up to 0.5%/εr with at most 6 μL of liquid in the channel.

A new inorganic-organic hybrid material produced from 2,6-dimethylanilinium cations and tin halide (SnCl₆)²⁻ has been synthesized and structurally determined by X-ray diffraction method. The title compound crystallizes in the monoclinic system, space group C2/m with a = 19.8772(4), b = 6.9879(1), c = 8.3001(2) Å, β = 98.487(2)° and V = 1140.26(4) ų. The crystal structure is built up of sheets of (SnCl₆)²⁻ octahedral anions and 2,6-xylidinium cations. The optical band gap was calculated and found to be 4.11 eV. At high temperature this compound exhibits a structural phase transition at 338 K. This has been characterized by differential scanning calorimetric and dielectric studies. Measurements of AC conductivity as a function of frequency at different temperatures indicated the hopping conduction mechanism. The bioassay results showed that the structure exhibits significant antibacterial activity.

Wrist-worn sensors have better compliance for activity monitoring compared to hip, waist, ankle or chest positions. However, wrist-worn activity monitoring is challenging due to the wide degree of freedom for the hand movements, as well as similarity of hand movements in different activities such as varying intensities of cycling. To strengthen the ability of wrist-worn sensors in detecting human activities more accurately, motion signals can be complemented by physiological signals such as optical heart rate (HR) based on photoplethysmography. In this paper, an activity monitoring framework using an optical HR sensor and a triaxial wrist-worn accelerometer is presented. We investigated a range of daily life activities including sitting, standing, household activities and stationary cycling with two intensities. A random forest (RF) classifier was exploited to detect these activities based on the wrist motions and optical HR. The highest overall accuracy of 89.6 ± 3.9% was achieved with a forest of a size of 64 trees and 13-s signal segments with 90% overlap. Removing the HR-derived features decreased the classification accuracy of high-intensity cycling by almost 7%, but did not affect the classification accuracies of other activities. A feature reduction utilizing the feature importance scores of RF was also carried out and resulted in a shrunken feature set of only 21 features. The overall accuracy of the classification utilizing the shrunken feature set was 89.4 ± 4.2%, which is almost equivalent to the above-mentioned peak overall accuracy.

This paper presents and discusses various metrics proposed for evaluation of polyphonic sound event detection systems used in realistic situations where there are typically multiple sound sources active simultaneously. The system output in this case contains overlapping events, marked as multiple sounds detected as being active at the same time. The polyphonic system output requires a suitable procedure for evaluation against a reference. Metrics from neighboring fields such as speech recognition and speaker diarization can be used, but they need to be partially redefined to deal with the overlapping events. We present a review of the most common metrics in the field and the way they are adapted and interpreted in the polyphonic case. We discuss segment-based and event-based definitions of each metric and explain the consequences of instance-based and class-based averaging using a case study. In parallel, we provide a toolbox containing implementations of presented metrics.

Public evaluation campaigns and datasets promote active development in target research areas, allowing direct comparison of algorithms. The second edition of the challenge on Detection and Classification of Acoustic Scenes and Events (DCASE 2016) has offered such an opportunity for development of state-of-the-art methods, and succeeded in drawing together a large number of participants from academic and industrial backgrounds. In this paper, we report on the tasks and outcomes of the DCASE 2016 challenge. The challenge comprised four tasks: acoustic scene classification, sound event detection in synthetic audio, sound event detection in real-life audio, and domestic audio tagging. We present in detail each task and analyse the submitted systems in terms of design and performance. We observe the emergence of deep learning as the most popular classification method, replacing the traditional approaches based on Gaussian mixture models and support vector machines. By contrast, feature representations have not changed substantially throughout the years, as mel frequency-based representations predominate in all tasks. The datasets created for and used in DCASE 2016 are publicly available and are a valuable resource for further research.

In this paper we propose a method for separation of moving sound sources. The method is based on first tracking the sources and then estimation of source spectrograms using multichannel non-negative matrix factorization (NMF) and extracting the sources from the mixture by single-channel Wiener filtering. We propose a novel multichannel NMF model with time-varying mixing of the sources denoted by spatial covariance matrices (SCM) and provide update equations for optimizing model parameters minimizing squared Frobenius norm. The SCMs of the model are obtained based on estimated directions of arrival of tracked sources at each time frame. The evaluation is based on established objective separation criteria and using real recordings of two and three simultaneous moving sound sources. The compared methods include conventional beamforming and ideal ratio mask separation. The proposed method is shown to exceed the separation quality of other evaluated blind approaches according to all measured quantities. Additionally, we evaluate the method's susceptibility towards tracking errors by comparing the separation quality achieved using annotated ground truth source trajectories.