1. Education
• February 2007 - PhD thesis from Université Joseph Fourier de Grenoble (UJF). Thesis: An Independent Component Analysis approach to the multicomponent image compression.
Supervisors: Dinh-Tuan Pham and Michel Barret.
• June 2003 - Master Degree in Engineering, Electronic and numerical telecommunication, with honours (mention Bien) from Ecole Nationale de l’Aviation Civile de Toulouse, ENAC, Toulouse, France.
• June 2003 - Master Degree, Signal, image and acoustic processing of Ecole Nationale Supérieure d' Electrotechnique, Electronique, d’Informatique, d' d’Hydraulique et des Telécommunications, ENSEEIHT, Toulouse, with honours (mention Bien). Supervisors: Cyril Dupouy and Christophe Macabiau. Thesis: System of GBAS (Global Positioning Based Augmentation System) interferences detection.
• July 2001 - Master Degree in Engineering, Electronic and telecommunication, with honours (mention Bien) from Ecole Nationale Suprieure Polytechnique de Yaoundé, ENSPY, Yaoundé, Cameroon.
2. Professionnal experience
• Sept. 2006 -August. 2007 - Teaching assistant in statistical inference and mathematical analysis at UJF, Grenoble, France.
• Nov. 2003 -Feb. 2007 - PhD thesis on independent component analysis (ICA) and multicomponent image compression at Laboratoire Jean Kuntzmann (Grenoble, France) and Supélec campus de Metz (Metz, France).
• Sept. 2004 -June. 2006 - Teaching assistant in statistical inference at UJF and Ecole Française de Papeterie Graphique de Grenoble (EFPG), Grenoble, France.
• Feb. 2003 -July. 2003 - Internship at Service Technique de la Navigation Aérienne (STNA) de Toulouse. Definition, conception and design of an electronic device for the detection of GBAS interference. We design an electronic device with a large band antenna associated with a low noise amplifier. We implemented a software using LABVIEW in order to continuously cover the wide spectrum concerned with a very low frequency. Before launching the use of GPS based system in civil aviation, it is necessary for civil authority aviation to ensure that those systems will always work with an excellent efficiency and efficacity, that was the purpose and the motivation of such studies.
• Jan. 2001 -July. 2001 - Internship at Laboratoire d’´electronique et de traitement du signal (LETS) de l’ENSPY de Yaoundé, Cameroon. Design of an electronic device
for EMG (Electromyogram) signal detection. Implementation in C++ of a software that made simple statistical analysis of EMG signal. Lossless compression of EMG signal with arithmetic and Huffman coding.
• July. 2000 -Sept 2000 - Internship at LETS de Yaoundé. Implementation of software for a VSAT antenna network. Evaluation of the characteristic between the VSAT and
the satellite in both up and down link. Calculation of the principal characteristics of the VSAT antenna. The software have been implemented in C++ and used by internet network providers.
June. 1999 -July 1999 - Internship of six weeks at Société Nationale d'électricité du Cameroun (SONEL), Ombé, Cameroon. The aim of the training was to learn the production, the transport and the distribution of the electrical energy. Most of the electrical energy in Cameroon is produced with hydroelectric dams.
• July. 1998 - Internship of five weeks at Société de Lingerie du Cameroun (SOLICAM) at Douala, Cameroon. Adaptation and integration in the working team of the electromechanical subdivision.
3. References
• Dinh-Tuan Pham, CNRS (Centre National de la Recherche Scientifique) Research director.
Tel: 33476514423,Email: Dinh-tuan.pham(at)imag.fr
• Michel Barret, Sup´elec assistant professor.
Tel: 33387764731,Email: Michel.barret(at)supelec.fr.
• Pierre Duhamel, CNRS Research director.
Tel: 33169851712,Email: Pierre.duhamel(at)lss.supelec.fr
• Catherine Lambert, CNES (Centre National d’Etudes Spatiales) Toulouse. Tel : 33 5 61 27 33 08, Email: Catherine.lambert(at)cnes.fr
• Serge Dégerine, UJF professor.
Tel: 33476635972,Email: Serge.Degerine(at)imag.fr
• Cyril Dupouy, Airbus research and development engineer.
Tel: 33561186804,Email: Cyril.Dupouy(at)airbus.fr
4. Computer skills
• Languages: - C, C++, Pascal.
• Statistical and simulation software - Matlab, Simulink, Labview.
5. Languages
• French: Mother tongue.
• English: Fluent. Writing, reading, speaking.
6. Teaching
• Teaching in second year at UJF university
Statistical inference tutorial class (108h).
.
The tutorial classes treat the bases of inference statistics: point estimation, confidence intervals, tests (on the mean, the variance, test of comparison of samples, Chi-2 test) and linear regression. The applications generally come from the biological background.
Statistical inference practical work class (90h).
Students have to get familiar with a software that can be used for simple statistical calculation (Excel or Calc of Open office). They have to use the different
statistical inference functions to solve concrete statistical problems. Students have to be able to do an elementary statistical study of a set of data. The practical work are related to the tutorial one, a report is done and evaluated for each class.
• Teaching in first year at UJF university
Mathematical analysis tutorial class (11h).
The tutorial classes treat the bases of mathematical analysis for engineer students. The lessons concern integration, multi variables function, differential equation and times series analysis.
• Teaching in first year EFPG (French engineering school) Statistics for engineer (12h).
Introduction to the inference statistics, punctual estimation and usual probability laws. The maximum likelihood is introduced to solve some concrete engineer problems.
7. Scientific activity summary and research project
Scientific activity summary
My research deals with multicomponent image compression and independent component analysis.
• Transform coding and independent component analysis The problem is to find optimal transformation for multicomponent images. Such images have a simple particularity with two kind of redundancy: the spatial one that exists between the pixels in each component and the spectral one that exists between
the components in the spectral axis. Recent previous work have shown that the discrete wavelet transform (DWT) allow good reduction of the spatial redundancy
of each component. For the spectral one, the Karhunen Loeve transform is usually used (with or without component reduction). However, this last transform is not the optimal one because the multicomponent image are not gaussian in the image domain or in the wavelet domain. In order to find the optimal spectral transform, we used the information theory assuming high rate coding.
Independent component analysis (ICA) is a new branch of the statistic domain and can be viewed as a particular case of the more global problem of blind source separation. The aim of ICA is to find sources from observations with a given modelling so that the sources are statistical independent. The mutual information is a criterium that allow to measure the dependence between random variables. If the random variables are independent then, their mutual information become equal to zero.
Using the information theory with high rate assumption, we show that the optimal linear spectral transform is obtained in some cases with the minimization of a criterion that is related to the one of ICA problem when the mutual information is the criterion to reduce. More precisely, we show that the compression criterion is equal to the criterion of the ICA added with a positive real term. This last term can be viewed as a pseudo distance to the orthogonality of the linear transform.
We have proposed different compression schemes and evaluated the performances of different spectral transform in those schemes. In image compression, one is usually concerned with the rate distorsion curve. We have used the verification model version 9 (VM9) of the JPEG2000 committee for the practical coding process of multispectral or hyperspectral images. We show that our new spectral transform performed better that the KLT that is usually used as the spectral transform.
Convolutive Transform
We investigate the introduction of a single transform that reduce both the spatial and the spectral redundancies. A convolutive transform is used for that purpose and we find out a criterion to be reduce for the compression of multicomponent image with high rate assumption coding. That criterion is related to the one that is used for independent component analysis assuming a convolutive modeling with the mutual information being the criterion to be minimized. We have minimized that criterion using a descent Newton algorithm.
Research project
I was essentially interested in multicomponent image compression during my PHD thesis, I have been focused on discrete wavelet transform, scalar and vector quantization and specific coders like SPIHT and EBCOT of JPEG2000 with image compression targeted as the main application. My researches were also focused on blind source separation, specially on independent component analysis.
My professional project
In the futur, I am interested to have a space career as Engineer or as a Research Engineer in the electronics and telecommunications fields associated with signal or image processing.
8. Publications, Conferences and Seminars
• Actual Working papers
Akam Bita Isidore Paul, Dinh-Tuan Pham, Multichannel Blind Deconvolution of Images Using Mutual Information Approach.
Akam Bita Isidore Paul, Michel Barret, Dinh-Tuan Pham, Compression of multicomponent images using an optimal separation deconvolution transformation.
Akam Bita Isidore Paul, Michel Barret, Dinh-Tuan Pham, Compression of multicomponent images using optimal linear transform.
Conferences and seminars
• Akam Bita Isidore Paul, Michel Barret, Dinh-Tuan Pham, Transformations linéaires optimales à hauts débits pour la norme JPEG2000 en compression d’images multicomposantes, GRETSI 2007.
Isidore Paul Akam Bita, Michel Barret et Dinh-Tuan Pham, Compression of Multicomponent Satellite Images Using Independent Component Analysis, Proceedings of the Sixth International Conference on Independent Component Analysis and Blind Signal Separation (ICA 2006), Charleston, Usa, vol. LNCS 3889, pp. 335–342, March 2006.
Michel Narozny, Michel Barret, Dinh-Tuan Pham and Isidore Paul Akam Bita, Modified ICA Algorithm for Finding Optimal Transforms in Transform Coding, IEEE 4th Int. Symp. on Image and Signal Processing Analysis, pp. 111–116, September 2005.
Isidore Paul Akam Bita, Michel Barret et Dinh-Tuan Pham, Compression of Multispectral Satellite Images Using Independent Component Analysis, Geometrical Transforms for Image Processing -Application to Satellite Image Restoration and Compression, Workshop, CNES (Centre National d’Etudes Spatiales), Toulouse, November 24, 2005.
Isidore Paul Akam Bita, Michel Barret et Dinh-Tuan Pham, A Separation Deconvolution Modelling for Multicomponent Images Compression, Transforms Based Independent Component Analysis, Workshop, TBICA 2006, Paris July 6-7, 2006.
Isidore Paul Akam Bita, Compression of Multicomponent Images Using Independent Component Analysis. Seminar SMS (Statistique et Mod´elisation Stochastique) at LJK (Laboratoire Jean Kuntzmann), Grenoble, December 1, 2005.
Isidore Paul Akam Bita, Optimal linear transform for multicomponent image with JPEG2000. Seminar at GIPSA-Lab (Laboratory of image, speech, signal processing and automatic of Grenoble), Grenoble, May 16, 2007.
Other works
• Akam Bita Isidore Paul, PhD thesis. Sur une approche de l’analyse en composantes indépendantes la compression des images multi composantes. February 12, 2007.
Akam Bita Isidore Paul, Master degree thesis. Système de détection des interfrences du GBAS. June 20, 2003.
Akam Bita Isidore Paul, Master engineering degree thesis. Acquisition et compression des signaux EMG. July 3, 2001.
Member of the projet ACI2M (Action concertée incitative masse de données) during three years. Coordinator of the project: Michel Barret (Michel.Barret(at)supelec.fr). |
Date Submitted:
