Ponències/Comunicacions de congressos
http://hdl.handle.net/2117/3926
2017-05-25T22:10:23ZSpatio-temporal road detection from aerial imagery using CNNs
http://hdl.handle.net/2117/104723
Spatio-temporal road detection from aerial imagery using CNNs
Luque, Belen; Morros Rubió, Josep Ramon; Ruiz Hidalgo, Javier
The main goal of this paper is to detect roads from aerial imagery recorded by drones. To achieve this, we
propose a modification of SegNet, a deep fully convolutional neural network for image segmentation. In
order to train this neural network, we have put together a database containing videos of roads from the point
of view of a small commercial drone. Additionally, we have developed an image annotation tool based on
the watershed technique, in order to perform a semi-automatic labeling of the videos in this database. The
experimental results using our modified version of SegNet show a big improvement on the performance of the
neural network when using aerial imagery, obtaining over 90% accuracy.
2017-05-22T16:13:24ZLuque, BelenMorros Rubió, Josep RamonRuiz Hidalgo, JavierThe main goal of this paper is to detect roads from aerial imagery recorded by drones. To achieve this, we
propose a modification of SegNet, a deep fully convolutional neural network for image segmentation. In
order to train this neural network, we have put together a database containing videos of roads from the point
of view of a small commercial drone. Additionally, we have developed an image annotation tool based on
the watershed technique, in order to perform a semi-automatic labeling of the videos in this database. The
experimental results using our modified version of SegNet show a big improvement on the performance of the
neural network when using aerial imagery, obtaining over 90% accuracy.Morphological interpolation for texture coding
http://hdl.handle.net/2117/104517
Morphological interpolation for texture coding
Casas Pla, Josep Ramon; Salembier Clairon, Philippe Jean; Torres Urgell, Lluís
In this paper a new morphological interpolation technique is presented. It is applied to the coding of the smooth (primary) component in a sketch-based image compression approach for very low bit-rates. The interpolation technique is intended to perform two dimensional interpolation from any set of initial pixels and, in particular, from sketch data. It makes intensive use of geodesic dilation, a morphological operator that may be implemented by means of FIFO queues. This results in a very efficient process compared to those that perform interpolation by linear filtering on the initial image. For the application of this method to interpolative image coding, the sketch data is extracted as a set of maximum curvature lines by means of the watershed algorithm. From such information, the interpolation technique obtains a fair reconstruction of both the smooth texture component and the main transitions of the image signal at low bit-rate cost.
2017-05-16T13:48:02ZCasas Pla, Josep RamonSalembier Clairon, Philippe JeanTorres Urgell, LluísIn this paper a new morphological interpolation technique is presented. It is applied to the coding of the smooth (primary) component in a sketch-based image compression approach for very low bit-rates. The interpolation technique is intended to perform two dimensional interpolation from any set of initial pixels and, in particular, from sketch data. It makes intensive use of geodesic dilation, a morphological operator that may be implemented by means of FIFO queues. This results in a very efficient process compared to those that perform interpolation by linear filtering on the initial image. For the application of this method to interpolative image coding, the sketch data is extracted as a set of maximum curvature lines by means of the watershed algorithm. From such information, the interpolation technique obtains a fair reconstruction of both the smooth texture component and the main transitions of the image signal at low bit-rate cost.Optical bistability in a fiber ring cavity with synchronous pulsed pump: anomalous and normal dispersion
http://hdl.handle.net/2117/104515
Optical bistability in a fiber ring cavity with synchronous pulsed pump: anomalous and normal dispersion
García Mateos, Jorge
Passive fiber cavities are basic nonlinear optical systems that are described by simple models - however they have a rich spectrum of complex behaviors (optical bistability, period-doubling bifurcations, chaos, modulational instability). From a practical point of view their study would have direct implications in the understanding of more complex cavity-based optical devices such as fiber lasers, Fabry-Perot lasers or APM lasers. Moreover, passive fiber cavities have potential applications in telecommunications as ultra-short pulse generators and as optical memories using their bistable behavior. This work is a contribution to the study of this bistable behavior with pulsed input, so that the device studied could be used for pulsed optical memory storage.
2017-05-16T13:21:36ZGarcía Mateos, JorgePassive fiber cavities are basic nonlinear optical systems that are described by simple models - however they have a rich spectrum of complex behaviors (optical bistability, period-doubling bifurcations, chaos, modulational instability). From a practical point of view their study would have direct implications in the understanding of more complex cavity-based optical devices such as fiber lasers, Fabry-Perot lasers or APM lasers. Moreover, passive fiber cavities have potential applications in telecommunications as ultra-short pulse generators and as optical memories using their bistable behavior. This work is a contribution to the study of this bistable behavior with pulsed input, so that the device studied could be used for pulsed optical memory storage.A new positive time-frequency distribution
http://hdl.handle.net/2117/104513
A new positive time-frequency distribution
Rodríguez Fonollosa, Javier; Nikias, C L
This article studies the formulation of new members of the Cohen-Posch (1985) class of positive time-frequency energy distributions. Members of this class are always positive functionals and satisfy the marginal constraints. Therefore, they can be properly interpreted as distributions. We considered the minimization of cross-entropy measures with respect to different priors or
2017-05-16T13:19:33ZRodríguez Fonollosa, JavierNikias, C LThis article studies the formulation of new members of the Cohen-Posch (1985) class of positive time-frequency energy distributions. Members of this class are always positive functionals and satisfy the marginal constraints. Therefore, they can be properly interpreted as distributions. We considered the minimization of cross-entropy measures with respect to different priors orMotion region overlapping for segmentation-based video coding
http://hdl.handle.net/2117/104512
Motion region overlapping for segmentation-based video coding
Pardàs Feliu, Montse; Salembier Clairon, Philippe Jean; Gonzalez, Benito
In object-based video coding systems, the scenes are described in terms of three dimensional objects, which can be coded as textures and contours. However, in order to achieve high compression ratios, the redundancy in the temporal dimension must be exploited using motion compensation. In this context, motion compensation can take advantage of the description of the scene as three dimensional regions. In this paper we present a motion estimation method which finds, in a fast and computationally simple manner, the affine parameters for every region of the scene. These parameters correspond to the motion of the texture. An ordering between regions is established in order to have the right compensation of contours.
2017-05-16T13:16:22ZPardàs Feliu, MontseSalembier Clairon, Philippe JeanGonzalez, BenitoIn object-based video coding systems, the scenes are described in terms of three dimensional objects, which can be coded as textures and contours. However, in order to achieve high compression ratios, the redundancy in the temporal dimension must be exploited using motion compensation. In this context, motion compensation can take advantage of the description of the scene as three dimensional regions. In this paper we present a motion estimation method which finds, in a fast and computationally simple manner, the affine parameters for every region of the scene. These parameters correspond to the motion of the texture. An ordering between regions is established in order to have the right compensation of contours.Application of hyperstability theory to interference cancelling
http://hdl.handle.net/2117/104511
Application of hyperstability theory to interference cancelling
Bertran Albertí, Eduardo; Montoro López, Gabriel
An alternative to the usual adaptive noise cancelling method devoted to removing interference is presented. In the conventional methodology to implement adaptive cancellers a reference signal is necessary correlated with the interference. This requirement (not always possible) is a limitation of this kind of canceller. The paper shows the use of model reference adaptive systems (MRAS), designed by using hyperstability theory, in order to cancel interference without the requirement of an auxiliary input. In the proposed methodology it is not necessary to have a reference signal to identify the interference; it is enough to know the upper and lower boundaries of this interference.
2017-05-16T13:13:02ZBertran Albertí, EduardoMontoro López, GabrielAn alternative to the usual adaptive noise cancelling method devoted to removing interference is presented. In the conventional methodology to implement adaptive cancellers a reference signal is necessary correlated with the interference. This requirement (not always possible) is a limitation of this kind of canceller. The paper shows the use of model reference adaptive systems (MRAS), designed by using hyperstability theory, in order to cancel interference without the requirement of an auxiliary input. In the proposed methodology it is not necessary to have a reference signal to identify the interference; it is enough to know the upper and lower boundaries of this interference.Continuous-time adaptive control applied to rf amplifier linearization
http://hdl.handle.net/2117/104507
Continuous-time adaptive control applied to rf amplifier linearization
Bertran Albertí, Eduardo; Montoro López, Gabriel; Talavera Pedrol, Núria
A new approach to the RF power amplifier linearization problem is presented. The proposed solution applies non-linear theories (Lyapunov direct method) to adaptive filtering in order to improve the linearity of the RF amplifiers. The obtained design requires lower circuit complexity than the LINC amplifier, and is not based on iterative algorithms nor sub-system identification. Up to 100 MHz these functions could be implemented, at present, with operational amplifiers and integrated analog multipliers (four quadrants). The adjusting algorithm convergence or the interruption of the communication are not problems in the proposed adaptive solution. The canceller structure design is based on model reference adaptive systems (MRAS): to cancel the error between the plant output (distortion output of the RF amplifier) and reference model (the desired signal obtained from a linear and low-power amplifier) by using continuous-time techniques. The proposed structure is studied by computer simulation (SPICE program) in a class-A RF power amplifier, The behaviour of the adapted amplifier is studied when power transistors approach nonlinear operating zones (saturation state).
2017-05-16T12:46:04ZBertran Albertí, EduardoMontoro López, GabrielTalavera Pedrol, NúriaA new approach to the RF power amplifier linearization problem is presented. The proposed solution applies non-linear theories (Lyapunov direct method) to adaptive filtering in order to improve the linearity of the RF amplifiers. The obtained design requires lower circuit complexity than the LINC amplifier, and is not based on iterative algorithms nor sub-system identification. Up to 100 MHz these functions could be implemented, at present, with operational amplifiers and integrated analog multipliers (four quadrants). The adjusting algorithm convergence or the interruption of the communication are not problems in the proposed adaptive solution. The canceller structure design is based on model reference adaptive systems (MRAS): to cancel the error between the plant output (distortion output of the RF amplifier) and reference model (the desired signal obtained from a linear and low-power amplifier) by using continuous-time techniques. The proposed structure is studied by computer simulation (SPICE program) in a class-A RF power amplifier, The behaviour of the adapted amplifier is studied when power transistors approach nonlinear operating zones (saturation state).Speaker recognition by means of restricted
Boltzmann machine adaptation
http://hdl.handle.net/2117/104488
Speaker recognition by means of restricted
Boltzmann machine adaptation
Safari, Pooyan; Ghahabi, Omid; Hernando Pericás, Francisco Javier
Restricted Boltzmann Machines (RBMs) have shown success in speaker recognition. In this paper, RBMs are investigated in a framework comprising a universal model training and model adaptation. Taking advantage of RBM unsupervised learning algorithm, a global model is trained based on all available background data. This general speaker-independent model, referred to as URBM, is further adapted to the data of a specific speaker to build speaker-dependent model. In order to show its effectiveness, we have applied this framework to two different tasks. It has been used to discriminatively model target and impostor spectral features for classification. It has been also utilized to produce a vector-based representation for speakers. This vector-based representation, similar to i-vector, can be further used for speaker recognition using either cosine scoring or Probabilistic Linear Discriminant Analysis (PLDA). The evaluation is performed on the core test condition of the NIST SRE 2006 database.
2017-05-16T10:10:01ZSafari, PooyanGhahabi, OmidHernando Pericás, Francisco JavierRestricted Boltzmann Machines (RBMs) have shown success in speaker recognition. In this paper, RBMs are investigated in a framework comprising a universal model training and model adaptation. Taking advantage of RBM unsupervised learning algorithm, a global model is trained based on all available background data. This general speaker-independent model, referred to as URBM, is further adapted to the data of a specific speaker to build speaker-dependent model. In order to show its effectiveness, we have applied this framework to two different tasks. It has been used to discriminatively model target and impostor spectral features for classification. It has been also utilized to produce a vector-based representation for speakers. This vector-based representation, similar to i-vector, can be further used for speaker recognition using either cosine scoring or Probabilistic Linear Discriminant Analysis (PLDA). The evaluation is performed on the core test condition of the NIST SRE 2006 database.NOMA and interference limited satellite scenarios
http://hdl.handle.net/2117/104486
NOMA and interference limited satellite scenarios
Caus, Marius; Vázquez, Miguel Ángel; Pérez Neira, Ana Isabel
This paper deals with the problem of non-orthogonal multiple access (NOMA) in multibeam satellite systems, where the signals are jointly precoded. It is considered that the number of frames that are simultaneously transmitted is higher than the number of feeds, reducing the precoding interference mitigation capabilities as the system becomes overloaded. In order to solve this problem, we assume that the satellite user terminals are able to perform multi-user detection to mitigate the interference. In the current NOMA approach, it
is assumed a successive interference cancellation (SIC) receiver.
To increase the spectral efficiency, this paper investigates NOMA
with simultaneous non-unique detection (SND). Compared to the
case where user terminals perform single user detection (SUD),
conventional scheduling heuristic rules do not longer apply in this
scenario. Therefore, different scheduling algorithms are proposed
considering both SIC and SND strategies. As the numerical evaluations show, SND yields larger average data rates than the SIC receiver. Concerning the scheduling, the best strategy is to pair users with highly correlated channels and the lowest channel gain difference. It is also shown that the sum-rate can be increased in overloaded satellite systems with respect to satellite scenarios, where the number of transmitted frames and feeds is the same.
2017-05-16T10:00:47ZCaus, MariusVázquez, Miguel ÁngelPérez Neira, Ana IsabelThis paper deals with the problem of non-orthogonal multiple access (NOMA) in multibeam satellite systems, where the signals are jointly precoded. It is considered that the number of frames that are simultaneously transmitted is higher than the number of feeds, reducing the precoding interference mitigation capabilities as the system becomes overloaded. In order to solve this problem, we assume that the satellite user terminals are able to perform multi-user detection to mitigate the interference. In the current NOMA approach, it
is assumed a successive interference cancellation (SIC) receiver.
To increase the spectral efficiency, this paper investigates NOMA
with simultaneous non-unique detection (SND). Compared to the
case where user terminals perform single user detection (SUD),
conventional scheduling heuristic rules do not longer apply in this
scenario. Therefore, different scheduling algorithms are proposed
considering both SIC and SND strategies. As the numerical evaluations show, SND yields larger average data rates than the SIC receiver. Concerning the scheduling, the best strategy is to pair users with highly correlated channels and the lowest channel gain difference. It is also shown that the sum-rate can be increased in overloaded satellite systems with respect to satellite scenarios, where the number of transmitted frames and feeds is the same.A study of flexible bandwidth allocation in statistical OFDM-based PON
http://hdl.handle.net/2117/104482
A study of flexible bandwidth allocation in statistical OFDM-based PON
Cano Valadéz, Iván Nicolás; Escayola Elias, Francisco Javier; Peralta Gallardo, Ángel; Polo Querol, Víctor; Santos Blanco, M. Concepción; Prat Gomà, Josep Joan
The upstream of an OFDM-PON with direct modulated lasers tuned in temperature was experimentally tested. Results demonstrate flexible bandwidth allocation, and show it constitutes an efficient strategy to cope with the differential link-loss among the users in the PON.
2017-05-16T09:51:15ZCano Valadéz, Iván NicolásEscayola Elias, Francisco JavierPeralta Gallardo, ÁngelPolo Querol, VíctorSantos Blanco, M. ConcepciónPrat Gomà, Josep JoanThe upstream of an OFDM-PON with direct modulated lasers tuned in temperature was experimentally tested. Results demonstrate flexible bandwidth allocation, and show it constitutes an efficient strategy to cope with the differential link-loss among the users in the PON.