Werner Creixell

Using RGB Image Processing for Designing an Alginate Edible Film

The use of edible films to coat food products is a technique that allows for an extended shelf-life. One of the most widely used polymers is calcium alginate. However, this polymer can modify the original food color and the perception by consumers. The objective was to design an alginate film based principally on color changes using a RGB color model. Edible films were prepared with sodium alginate and glycerol as plasticizer, cross-linking the polymer with calcium. Dry and hydrated states of the edible films were studied. Film thickness was directly proportional to surface concentration and increased with hydration. There is a zone in which the color does not change with alginate surface concentration and another where the color is directly proportional to it. This latter scenario is not a consequence of structural changes or the degree of hydration. Results showed a range where the color was not modified by the alginate concentration; hence, an optimal surface concentration was determined as a design parameter. Edible films made using the optimal surface concentration would not mask microbial contamination and have good physical properties (water vapor transmission and swelling) compared with other surface concentrations. In addition, it was possible to model alginate surface concentration as a function of surface color using mathematical tools (clustering, linear regression, and support vector machine), allowing one to study the optimal use of the edible films.

Modeling Volatile Organic Compounds Released by Bovine Fresh Meat Using an Integration of Solid Phase Microextraction and Databases

The influence of volatile organic compounds is often regarded as a critical factor of food flavor quality. Nevertheless, information about the biochemical origin of odorants is not totally available. Quantification of volatile compounds in meat is not easy, but a new approach that helps is the use of GC/MS–SPME. This technique can generate large-scale data sets that allow for the development of computational technologies for better integration and biochemical interpretation. Volatile organic compounds of bovine fresh meat samples were measured by GC/MS–SPME, using four SPME fibers. The compounds were analyzed using an integration of chemical and biological database systems (NIST and KEGG, respectively). Twelve metabolic pathways were identified. A metabolic network model was constructed using graph elements, linking pathways through a central branch mediated by glycolysis. Experimental data indicated that post-mortem glycolysis is associated with the release of volatile compounds in fresh meat. The proposed technique could be used to study possible sources of biochemical compounds in meat to be applied in food analysis.

Mobility prediction algorithm for mobile ad hoc network using pedestrian trajectory data

In the last years mobile ad hoc networks (MANET) has became a hot research subject. MANET is compound by mobile nodes that can be carried by people; the network formed by these nodes has then a highly changing topology. Nodes move according to people mobility patterns and a big number of the network tasks are influenced by nodes mobility. It is reasonable to think that network performance is also highly influenced by mobility. Reduce the uncertainty would have a positive impact on networks performance. In this work we propose a method to dynamically predict a pedestrian position. The method uses recursive least squares lattice adaptive filter to achieve that goal with reasonable low error.

Meta-learning based optimization of metabolic pathway data-mining inference system

This paper describes a novel meta-learning (MTL) based methodology used to optimize a neural network based inference system. The inference system being optimized is part of a bioinformatic application built to implement a systematic search scheme for the identification of genes which encode enzymes of metabolic pathways. Different MTL implementations are contrasted with manually optimized inference systems. The MTL based approach was found to be flexible and able to produce better results than manual optimization.

Serendipity in social networks

Serendipity is defined as fortunate discoveries made by chance. In this work we explore the idea that topological measures of a persons social network could be an indicator about how likely that person is to experience fortunate discoveries.

EM-based ML channel estimation in OFDM systems with phase distortion using RB-EKF

In this paper we address the joint estimation of the channel impulse response in orthogonal frequency division multiplexing systems with phase distortion, namely phase noise and carrier frequency offset, phase noise bandwidth and the additive noise variance. The estimation algorithm is based on an implementation of the Extended Kalman Filter within the general framework of the Expectation-Maximization algorithm. We focus on the partial training case, where the transmitted signal is not fully known. To tackle this problem, we utilize a Rao-Blackwellized Extended Kalman Filter. We also compare our results with another nonlinear filtering technique, namely Rao-Blackwellized Particle Filtering, applied to this joint estimation problem. The performance of the two filtering techniques considered in this paper is evaluated in terms of the mean square error of the channel estimates and the numerical complexity introduced by each of these techniques.

Recursive parameter and state estimation for a mining industry process

In this paper we consider the problem of recursive state and parameter estimation of a nonlinear system. We propose an approach where we combine Particle Filter (for state estimation) and Recursive Least Squares (for parameter estimation) The class of nonlinear systems is motivated by a real process of the copper mining industry. The proposed approach is tested with real data.

Automatic cross-language plagiarism detection

In this paper a new method to detect cross-language plagiarism in electronic documents is presented. The method is based on natural language processing techniques and machine learning methods. Preliminary results show that the system is able to identify the Internet sources of plagiarism of texts written in Spanish that have been copied (by humans and machines) from English sources.

Routing Protocol for Mobile Ad HOC Networks Using Mobility Prediction

Over the lasts years there has been a growing interest on mobile ad hoc networks (MANET), in this network nodes are carried by people and connected to each other by multi-hop wireless links. Since the nodes can move freely, then the nodes mobility pattern became an important characteristic of the MANET. In this work we proposed a geographical routing protocol which takes forwarding decisions based on the current nodes position and the its future position. To estimate the future position of the node we used a prediction method who has been based on real trajectories data. The algorithm proposed is compared with other proposed algorithms by simulation

Tracking and following a moving object with a quadcopter

This work implements a tracking algorithm using video streamed from a flying platform in order to follow a moving object. The platform consists of a low cost quadcopter with a Wi-Fi camera which sends compressed H.264 video to a ground station. The object to be followed was defined by a bounding box on a single frame. In order to identify the object, each video frame was processed using motion estimation and a cascaded classifier. For the movement control a proportional integral controller (PI) was used. Its inputs were detection size, out of plane rotation and estimated object position. The controller outputs were pitch, yaw, roll and altitude respectively. Each movement compensation was computed by independent controllers. The orientation commands were sent back to the quadcopter closing the loop. The proposed framework was tested by following a person while walks outdoor. Experiments showed remarkable performance indicating this proposal is useful for surveillance tasks.