Search

This paper describes a novel DNN-based system, named PD3net , that detects multiple people from a single depth image, in real time. The proposed neural network processes a depth image and outputs a likelihood map in image coordinates, where each detection corresponds to a Gaussian-shaped local distribution, centered at each person’s head. This likelihood map encodes both the number of detected people as well as their position in the image, from which the 3D position can be computed. The proposed DNN includes spatially separated convolutions to increase performance, and runs in real-time with low budget GPUs. We use synthetic data for initially training the network, followed by fine tuning with a small amount of real data. This allows adapting the network to different scenarios without needing large and manually labeled image datasets. Due to that, the people detection system presented in this paper has numerous potential applications in different fields, such as capacity control, automatic video-surveillance, people or groups behavior analysis, healthcare or monitoring and assistance of elderly people in ambient assisted living environments. In addition, the use of depth information does not allow recognizing the identity of people in the scene, thus enabling their detection while preserving their privacy. The proposed DNN has been experimentally evaluated and compared with other state-of-the-art approaches, including both classical and DNN-based solutions, under a wide range of experimental conditions. The achieved results allows concluding that the proposed architecture and the training strategy are effective, and the network generalize to work with scenes different from those used during training. We also demonstrate that our proposal outperforms existing methods and can accurately detect people in scenes with significant occlusions.

One of the most popular techniques for Acoustic Source Localization is the Generalized Cross Correlation (GCC) and its use in Steered Response Power techniques (SRP). Nowadays, Deep Learning strategies may outperform these classical methods, but they are generally dependent on the room and sensor geometric configuration that are used during the training phases. Hence, they require adaptation and re-training when facing a new environment, which is a problem in practice as re-training requires labelling new data and running a complex training algorithm. In this work we use a Convolutional Deep Neural Network that transforms the GCC between two signals into a Gaussian shaped signal, that we call Deep Generalized Cross Correlation (DeepGCC). We combine DeepGCC estimations to create a 3D acoustic map, similarly to SRP techniques. This acoustic map can be further refined using a sparse generative model to recover the source position. Crucially, we can adapt the acoustic map to different microphone array geometries without retraining the DeepGCC network. We show that our method outperforms both classical approaches and recent Deep Learning strategies in real and simulated challenging scenarios with mismatched training-testing conditions, not requiring re-training with different sensor configurations or room environments.

The algorithms for acoustic source localization based on PHAT filtering have been profusely used with good results in reverberant and noisy environments. However, there are very few studies that give a formal explanation of their robustness, most of them providing just an empirical validation or showing results on simulated data. In this work we present a novel analytical model for predicting the behavior of both the SRP-PHAT power maps and the GCC-PHAT functions. The results show that they are only affected by the signal bandwidth, the microphone array topology, and the room geometry, being independent of the spectral content of the received signal. The proposed model is shown to be valid in reverberant environments and under far and near field conditions. Using this result, an analysis study on how the aforementioned factors affect the SRP-PHAT power maps is presented providing well supported theoretical and practical considerations. The model validation is based on both synthetic and real data, obtaining in all cases a high accuracy of the model to reproduce the SRP-PHAT power maps, both in anechoic and non-anechoic scenarios, becoming thus an excellent tool to be exploited for the improvement of real world relevant applications related to acoustic localization. HighlightsA novel parametric analytical model to predict SRP-PHAT power maps is formulated.An exhaustive evaluation is done on both synthetic and real data.Results show high accuracy for very different acoustical and geometrical conditions.The paper also addresses practical issues in the model implementation.

A specific study of the orthographic, weak-perspective and paraperspective cameras.An algebraic procedure solving the affine correction problem for each camera model.A closed-form solution for each camera model.A full characterization of the affine corrections' generic ambiguities for each camera model.An experimental evaluation comparing the algebraic procedures to global polynomial optimization and an interior-point method. Given a general affine camera, we study the problem of finding the closest metric affine camera, where the latter is one of the orthographic, weak-perspective and paraperspective projection models. This problem typically arises in stratified Structure-from-Motion methods such as factorization-based methods. For each type of metric affine camera, we give a closed-form solution and its implementation through an algebraic procedure. Using our algebraic procedure, we can then provide a complete analysis of the problem's generic ambiguity space. This also gives the means to generate the other solutions if any.

One of the leading time of flight imaging technologies for depth sensing is based on Photonic Mixer Devices (PMD). In PMD sensors each pixel samples the correlation between emitted and received light signals. Current PMD cameras compute eight correlation samples per pixel in four sequential stages to obtain depth with invariance to signal amplitude and offset variations. With motion, PMD pixels capture different depths at each stage. As a result, correlation samples are not coherent with a single depth, producing artifacts. We propose to detect and remove motion artifacts from a single frame taken by a PMD camera. The algorithm we propose is very fast, simple and can be easily included in camera hardware. We recover depth of each pixel by exploiting consistency of the correlation samples and local neighbors of the pixel. In addition, our method obtains the motion flow of occluding contours in the image from a single frame. The system has been validated in real scenes using a commercial low-cost PMD camera and high speed dynamics. In all cases our method produces accurate results and it highly reduces motion artifacts. Display Omitted Motion Artifacts (MoArt) are a non-systematic error in Time of Flight (ToF) imaging.MoArt emerge when motion appears during the integration time.MoArt cause distorted ToF measurements (depth and amplitude).We propose a single frame correction of ToF measurements with motion flow estimation.We present quantitative and qualitative results in challenging scenarios.

Multipath interference of light is the cause of important errors in Time of Flight (ToF) depth estimation. This paper proposes an algorithm that removes multipath distortion from a single depth map obtained by a ToF camera. Our approach does not require information about the scene, apart from ToF measurements. The method is based on fitting ToF measurements with a radiometric model. Model inputs are depth values free from multipath interference whereas model outputs consist of synthesized ToF measurements. We propose an iterative optimization algorithm that obtains model parameters that best reproduce ToF measurements, recovering the depth of the scene without distortion. We show results with both synthetic and real scenes captured by commercial ToF sensors. In all cases, our algorithm accurately corrects the multipath distortion, obtaining depth maps that are very close to ground truth data.

This paper tackles the problem of identification and tracking of multiple robots in an intelligent space using an array of cameras placed in fixed positions within the environment. Several types of agent can be found in an intelligent space: controlled agents (mobile robots) and uncontrolled ones (users and obstacles). The information transferred between the controlled agents and the intelligent space is limited to the control commands sent to the robots and the measurements of the odometers received from the robots. The proposed solution allows the localization of mobile agents, even if they are not robots; however, we have focused on the controlled agents. The proposal does not require prior knowledge or invasive landmarks on board the robots. It starts from the segmentation of different agents in motion that allows obtaining the boundaries of all robots and an estimation of all 3D points that define those boundaries. Then, the identification of the robots is obtained by comparing the components of the linear velocity estimated by the motion segmentation algorithm and received from the odometers. In order to track the robots, an eXtended Particle Filter with Classification Process (XPFCP) is employed. Several experimental tests have been carried out, and the results obtained validate the proposal.

Thirty-five infants with hypernatremic diarrheal dehydration were given "slow" oral rehydration therapy, with deficits replaced over a period of 12 hours. A group of 24 infants received glucose-electrolyte solution for 8 hours, followed by plain water for 4 hours in a volume of 2:1; 11 other infants received equivalent volumes of glucose-electrolyte solution alone over 12 hours. Serum sodium concentrations fell to normal at similar rates in both groups. None of the 35 infants manifested convulsions. These preliminary results indicate that further evaluation of slow oral rehydration in infants with hypernatremic dehydration should be considered.

artículo -- Universidad de Costa Rica, Instituto de Investigaciones en salud. 1983. Este documento de privado debido a limitaciones de derechos de autor. Two hundred thirty-four of 242 neonates (96.7 %) with a mean dehydration of 5.1% body weight were rehydrated with a glucose-electrolyte solution given orally in an average time of 7 44 hours. Eight (3.3 %) neonates required intravenous fluid therapy. Emesis was not an obstacle for complete oral rehydration. Hypernatremia, hyponatremia, and acidosis present on admission were corrected within a few hours. Oral rehydration is an entirely satisfactory choice for treatment of uncomplicated dehydration in neonates. Universidad de Costa Rica, Instituto de investigaciones en Salud UCR::Vicerrectoría de Investigación::Unidades de Investigación::Ciencias de la Salud::Instituto de Investigaciones en Salud (INISA)

PIP: Doctors Kahn and Blum based their views on oral rehydration on only 7 cases, and they fail to provide their methodological details. In their letter on oral rehydration with UNICEF/WHO (United Nations International Children's Emergency Fund/World Health Organization) glucose electrolyte solution (GES), they maintain that hyperkalemia is a danger of GES therapy, that hypernatremia will be aggravated, that therapy should not last for longer than 24 hours, that after 24 hours monitoring of plasma potassium will be needed, and that except for developing countries where material milk is used, no plan of treatment has been proposed after the first 24 hours of rehydration. The experience of Kahn and Blum is at variance with extensive data from many carefully monitored balanced studies in infants treated with GES. GES is a potent medication and needs to be used properly. Guidelines for use are listed. Kahn and Blum fail to indicate whether their 7 patients comprised their entire treatment group or only those with biochemical or clinical problems. They also fail to indicate the degree of dehydration of the infants at onset of therapy or the extent of ongoing diarrheal losses, and they do not describe the precise treatment regimen. Their mean time of treatment -- 41 hours -- was particularly long. The hyperkalemia reported by Kahn and Blum may have resulted from excessive GES administration, without a source of free water, to infants having few diarrheal stools. Proper use of GES formula rapidly rehydrates 95-98% of mildly to severely dehydrated infants, irrespective of etiology.