Target tracking with template matching filtering

Published in 24 Thursday

Track

Pattern recognition

An efficient algorithm for target tracking based on template matching filtering is presented. The algorithm is able track the position of a target with invariance to appareance deformations, occlusions, and clutter.The target is defined by the user at the beginning of the algorithm. Next, a composite filter is designed torecognize the target in the next frame. The filter is adapted for each frame using information of currentand past scene frames. Results obtained with the proposed algorithm using real-life scenes, are presented and compared with those obtained with a recent state-of-the-art tracking method, in terms of detection efficiency, tracking accuracy, and speed of processing.

Read more...

Face recognition with correlation filters designed with multi-objective combinatorial optimization

Published in 24 Thursday

Track

Pattern recognition

A reliable approach for face recognition using composite correlation filters is presented. The filters are designed by combining several face images which are chosen by means of multi-objective combinatorial optimization. Given a vast search space of available face images, an iterative algorithm is used to synthesize a filter bank with an optimized performance in terms of several performance metrics. Computer simulation results obtained with the proposed method for face recognition in noisy scenes are discussed and compared in terms of recognition performance with existing state-of-the-art methods.

Read more...

EEG signal implementation of Movement Intention for the Teleoperation of the Mobile Differential Robot

Published in 24 Thursday

Track

Pattern recognition

In the year 1929 a German psychiatrist, named Hans Berger, demonstrated for the first time that the electric activity of the human brain was related to the person’s mental state. He also announced the possibility of registering such type of electric activities without opening the human head, i.e. noninvasive procedure, and that such electric activities could be plotted on a graph. Berger called such type of registration as electroencephalogram (EEG). EEG signals research has been growing over the years due to the their increasing use to control electronic devices in all sorts of contexts. The present work developed a prototype to control a differential robot by means of EEG signals using the detection of movement intention of the right and left hand. The study covered on one hand, the analysis and design of the teleoperation system, and on the other hand, the robot tele operational tests. It is important to point out that the robot was designed and built to meet the technical research purposes.The programming of the EEG signal processing was made using the API provided by MATLAB. In turn, the programming for controlling the mobile differential robot was made with Wiring and Python.Lastly, several tests and experiments were carried out, and they showed that the objective in view was met.

References

[1] L. Gutierrez-Flores, C. Aviles-Cruz, J. Villegas-Cortez, and A. Ferreyra-Ramirez. EEG PATTERNRECOGNITION: Application to a Real Time Control System for Android-Based Mobile Devices.Lecture Notes in Computer Science, volume 7914, pages 232–241. Springer Berlin Heidelberg, 2013.

[2] J.I. Perez Arregun, S. Tovar Arriaga, U.G. Villasenor Carrillo, E. Gorrostieta Hurtado, J.C. PedrazaOrtega, J.E. Vargas Soto, J.M. Ramos Arregun, A. Sotomayor Olmedo. Robot móvil de tracción difer-encial con plataforma de control modular para Investigación y desarrollo ágil de proyectos. 10 CongresoNacional de Mecatrónica, Asociación Mexicana de Mecatrónica A.C. Noviembre 3, 2011. Puerto Vallarta Jalisco.

[3] C. Villanueva Escudero, J. Villegas Cortez, A. Zuniga López, and C. Avilés Cruz. Monocular Visual Odometry Based Navigation for a Differential Mobile Robot with Android OS. Human-Inspired Com-puting and Its Applications - Lecture Notes in Computer Science, vol. 8856, pp. 281-292. Springer International Publishing, 2014.

Read more...

Human face classification by means of a local texture analysis using the CBIR technique and Points of Interest

Published in 24 Thursday

Track

Pattern recognition

The recognition of human faces represents an ongoing and very active area of research. This interest derives from the challenges posed by illumination occlusions and temporality. On the other hand, itsapplications continue to be very important, and more oriented toward security. During this lecture I am going to present a tested methodology for human faces classification on the basis of the analysis of the local texture of the face and contemplating the points of interest and the Content–Based Image Retrieval (CBIR) technique. The results achieved are excellent and the challenges lying ahead are of great interest, both for numerical floating point computing and Big Data applications.

References

[1] Gonzalo Pajares M. and Jesus M. de la Cruz G. Visión por computador imágenes digitales y aplicaciones. Alfa omega, 2da edición, Marzo 2008.

[2] J. F. Serrano-Talamantes, C. Aviles-Cruz, J. Villegas-Cortez, and J. H. Sossa-Azuela, Self organizing natural scene image retrieval, Expert Systems with Applications, 2012.

[3] K–A–Kim. Facial feature extraction using PCA and wavelet multi–resolution images. Sixth IEEE International Conference on Automatic Face and Gesture Recognition, pages 17-19, Mayo 2004.

[4] H.Bay, A.Ess, T. Tuytelaars, and L.V–Goo. Surf: Speeded up robust features. Computer Vision and Image Understanding (CVIU), 110(3):346-359, 2008.

[5] Krystian Mikolajczyk and Cordelia Schmid. A performance evaluation of local descriptors. IEEE Trans.Pattern Anal. Mach. Intell., 27(10):1615-1630, October 2005.

[6] Javier Ruiz-del Solar, Rodrigo Verschae, and Mauricio Correa. Recognition of faces in unconstrained environments: A comparative study. EURASIP J. Adv. Signal Process, 2009:1:1-1:19, January 2009.

Read more...

LSA.Studio: Augmenting the LSA technique in pervasive environments

Published in 25 Friday

Track

Pattern recognition

Ubicomp researchers have given the importance to the use of the technique LSA (Lag Sequential Analysis)[8, 7] to evaluate the impact of their systems. LSA is a technique for gathering quantitative data by observing users as they perform their normal activities. It is traditionally used in the field of developmental psychology to study the behavior of person to person interaction by measuring the number of times certain behaviors precede or follow a selected behavior; the behaviors are defined by the study evaluators. Datacan be captured live with paper and pencil or coded from video. Recent work of pervasive environments[2, 1] and social sciences [4, 6] shows that LSA is widely used for the analysis and evaluation, since it allows to capture the natural work of the user, also allows to obtain quantitative and statistical results from the data encoded.

An advantage of LSA is that it is conducted in the users environment, and it is conducted while the user performs his normal activities. With LSA, evaluators can generate statistics that capture aspects of observed behavior such as frequency and conditional probabilities of events. If video is used to capture the data, it can be recoded for different information as evaluation needs change, and it can be used for qualitative observational purposes. A significant disadvantage of LSA is cost; coding video for LSA is time consuming. There are many advantages for the use of LSA however there is a need to develop strategies to reduce the cost of its use [2].

This paper presents a tool that uses state of art computer vision techniques that allows the user to automate the process of encoding in LSA. Our aim is to augment LSA, reducing its burden, providing an specialized LSA. The tool presents a graphical interface for the user. How the tool works can be divided into four main steps: load video, select reference image, tracking process and classification. In the load video, the user manually selects a video. In the next main step, the user manually selects the image of the video where the behavior of interest is present. Then, a composite filter correlation [3] is used to process the reference image. At this point, the tool is ready to track the reference image in the video, this is done automatically by the tracking process; the TLD (Tracking Learning Detector) [5] tracking algorithm in computer vision is used for this process. When the TLD algorithm finds a possible match, the correlation peak threshold is used to determine a positive match between the processed reference image and possible image match.Finally, a time-stamp is registered at every positive match indicating that a behavior of interest occurred in that moment.

References

[1] Karina Caro. Exergames for children with motor skills problems. SIGACCESS Access. Comput.,(108):20–26, January 2014.

[2] S Consolvo, L Arnstein, and B R Franza. User study techniques in the design and evaluation of aubicomp environment. InIn proceedings of the 4th international conference on Ubiquitous Computing, 2002.

[3] Leopoldo. Gaxiola, Víctor Hugo Díaz-Ramírez, JuanJ. Tapia, Arnoldo Diaz-Ramirez, and Vitaly Kober.Robust face tracking with locally-adaptive correlation filtering. In Eduardo Bayro-Corrochano and Edwin Hancock, editors,Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, volume 8827 ofLecture Notes in Computer Science, pages 925–932. Springer International Publishing, 2014.

[4] Philip L. Gunter, Susan L. Jack, Richard E. Shores, Debra E. Carrell, and Julia Flowers. Lag sequentialanalysis as a tool for functional analysis of student disruptive behavior in classrooms.Journal ofEmotional and Behavioral Disorders, 1(3):138–148, 1993.

[5] Zdenek Kalal, Krystian Mikolajczyk, and Jiri Matas. Tracking-learning-detection.IEEE Trans. PatternAnal. Mach. Intell., 34(7):1409–1422, July 2012.

[6] Florian E. Klonek, Nale Lehmann-Willenbrock, and Simone Kauffeld. Dynamics of resistance to change:A sequential analysis of change agents in action.Journal of Change Management, 14(3):334–360, 2014.

[7] G P Sackett. Lag sequential analysis as a data reduction technique in social interaction research. InIn, pages 300–340. Brunner/Mazel, 1980.

[8] Gene P. Sackett, Richard Holm, Charles Crowley, and Allen Henkins. A fortran program for lag sequential analysis of contingency and cyclicity in behavioral interaction data.Behavior Research Methods and Instrumentation, 11(3):366–378, 1979.

Read more...

Profiting from several recommendation algorithms using a scalable approach

Published in 24 Thursday

Track

Pattern recognition

1 Introduction

Internet has become the biggest world market: thousand of companies reach every connected home offering millions of products to global customers, and recommender systems (RSs) are required to present products that may be of interest to the users. Although different approaches have been applied to develop RSs in the last decade [1][2], we consider in this paper the possibility of using several algorithms instead of a single one, trying to produce better recommendations, which requires high performance computing platforms to run all of the algorithms together. The system response must be then selected among those provided by each of the RSs responses. We apply a Fuzzy Rule-Based System (FRBS) to take the decision required.The parameters used by the FRBS are calculated from historical data from users and items.The main problem is the time required to compute each of the RSs every time a new recommendation must be generated. We decided to employ a Big Data approach relying on the well known Hadoop framework[3] which allows to parallelize the task, thus shortening execution time.

2 Methodology

Our methodology relies in two processes, an offline and online process. The offline process, which is split and run as a number of Hadoop jobs, generates user profiles which are used in the online process to decide what recommendation algorithm should generate the final recommendation.Several parameters are calculated by this offline process to obtain the necessary knowledge from users and items. The number of items that each user has rated are the basis for computing user profiles and parameters required, which are later employed to feed the FRBS in charge of selecting the final output of the system among those provided by each of the RSs.for the FRBS, it enables us to obtain the best recommendation depending on the user profile.

3 Results

In order to test the scalability of the proposal, we have run the recommender system using different number of computing nodes in a hadoop based cluster. As we notice in figure 1, the system shows scalability.Along the tests, the machine resources have been monitored. We noticed that CPU usage has been equally distributed along the machines, and the memory usage was not high. Reading and writing phases of Hadoop jobs were done in local disks, instead of using remote disks, leading to lower network traffic.In contrast, HBase scalability was not as expected. We detected that when several clients are writing into HBase, adding new nodes does not notably improve the performance. However, Hadoop jobs that reads from HBase data have a high performance, obtaining a very good scalability.

4 Conclusions

The solution presented in this paper has shown the benefits of using Hadoop for distributing and parallelizing a recommender system where different recommendation algorithms are joined by an FRBS.

5 Acknowledgments

Spanish Ministry of Economy, Project UEX:EPHEMEC (TIN2014-56494-C4-2-P) and CDTI project SmartCity Platform; Gobex, FEDER GRU10029.

References

[1] Aksel, F., Birtürk, A. Enhancing Accuracy of Hybrid Recommender Systems through Adapting theDomain Trends. In Workshop on the Practical Use of Recommender Systems, Algorithms and Tech-nologies (PRSAT 2010) (p. 11).

[2] M. Garcia-Valdez and A. Alanis. Fuzzy inference for Learning Object Recommendation. Fuzzy Systems(FUZZ), 2010.

[3] Hadoop, distributed scalable fault-tolerance framework for data processing. http://hadoop.apache.org/

[4] HBase, distributed and scalable database. http://hbase.apache.org/

[5] Mahout, machine learning library for Big Data solutions. http://mahout.apache.org/

Read more...
Subscribe to this RSS feed