Non-linear blind signal separation for chemical solid-state sensor arrays

• Autores: Guillermo Bedoya Jimenez
• Directores de la Tesis: Sergio Bermejo Sánchez (dir. tes.)
• Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2006
• Idioma: español
• Tribunal Calificador de la Tesis: Augustin Martinez (presid.), Juan Manuel Moreno Arostegui (secret.), Pierre Temple (voc.), Christian Jutten (voc.), Jordi Solé Casals (voc.)
• Materias:
  • Matemáticas
    • Estadística
      • Series temporales
  • Ciencias tecnológicas
    • Tecnología electrónica
    • Tecnología de la instrumentación
      • Instrumentos electrónicos
• Texto completo no disponible (Saber más ...)
• Resumen

  • This work is concerned with the problem of Nonlinear Blind Source Separation (BSS), In a more specific context, we deal with the implementation of a new class of smart sensor systems founded on the use of semiconductor-based chemical sensor arrays, interface electronics and neural-based source separation algorithms. The main goal is to obtain low cost/high performance integrated sensor system architectures. In the particular case of chemical sensors with poor performance, we propose the use of novel extended Nonlinear BSS techniques to improve their performance, give spatial selectivity and cancel cross-talks between several sources in the array. The thesis is focused on the search of the increase in efficiency and performance of smart chemical sensor systems that includes noise-tolerant and non-linear signal processing tools, with the role of allows a signal enhancement for the extraction of useful information from the sensors output. This implies researching basic questions such as what are strategies that can optimally use the information contained in the data from the statistics perspective, and the study, extension and application of particular Nonlinear BSS mixtures models.

    The main work is focused on the next topics:

    1. Study and development of source separation algorithms based on the post nonlinear mixture model, and its application to semiconductor-based bio-chemical sensor arrays.

    2. Study and development of source separation algorithms based on the bilinear mixture model, and its application to TGS gas sensor arrays systems.

    3. Development of signal processing strategies to improve the Bio-chemical and Gas sensors performance, given spatial selectivity and cancelling non linearities and cross-talksbetween several species in an array. Development of advanced learning strategies oriented to separate single chemical species from multi-component mixtures in high-noise environments.

    4. A global description of the algorithmic constraints and future