Research Statement

A major goal of computer engineering is to better immerse computers into human everyday activities, so that computer technology may improve the inter-relation among humans, computers, and the environment. The past decade has brought an enormous improvement in microprocessor and memory technologies, but, unfortunately, these improvements have not been reflected in the interfaces between humans and machines. Current interfaces (keyboards, mice, sliding knobs, graffiti stylus) are inconvenient, inadequate, unnatural, and difficult to use. None of the current interfaces are suitable to interact with unconventional, embedded computer systems such as wearable computers, video tape recorders, automobiles, and robots. In fact, current interfaces were designed for habitual computer users and for a limited range of tasks. Advances in VLSI, electronics, MEMS, and nanotechnology have dramatically decreased the prices of sensors (cameras, microphones, infrared, and others). These sensors are now the natural candidates around which better human-machine interfaces should be built.

My research focuses broadly on the interaction between humans and computers, a key component of such immersion. I seek to develop multi-modal, intuitive, and user-friendly human-machine interfaces that lead to a richer and more natural human-machine interaction. My research goals are two-fold. I am interested in the theoretical development and the formal characterization of novel algorithms for human-machine interaction. I am also very keen on the practical implementation of such algorithms and on the transfer of technology into actual industrial applications.

I have worked in different aspects of multimodal human-machine interaction for embedded systems. During my PhD thesis, I designed and built a computer interface that captures handwriting using a single camera. The user was allowed to write at will on a normal piece of paper with a common pen while, at the same time, the handwritten text was being captured. I further developed a personal identification system using signatures captured with the interface. The performance of the system was shown to be comparable to the best performances of camera-based identification systems present in the literature. Upon completion of my PhD, I joined VocalPoint Technologies, where I participated in the development of robust speech recognition algorithms for telephony.

After VocalPoint, I joined Evolution Robotics, a technology company focused on developing a software platform for robotics applications, where I have led the human-robot interaction group. A robot needs to interact directly with a noisy, dynamic, and unpredictable environment. Thus, the design of the human-robot interaction is vital for creating an interesting and engaging relationship between humans and robots. At Evolution Robotics, I have developed human-robot interaction systems based on vision and speech. I have also worked on a novel interaction mode for the Sony AIBO robot that is based on a very efficient and robust visual pattern recognition system that was ported (and optimized) from a Pentium-class CPU to an embedded platform.

The area of visual pattern recognition and its application to visual simultaneous localization and mapping (visual SLAM) has been my primary focus lately. My group of R&D at Evolution Robotics has been working enhancing the illumination independence of the recognition algorithm. We have also refined the visual SLAM system to handle full 6DOF motion.

For more information on the mentioned projects visit the following pages:

My research interest include:

  • Computer vision
  • Human-machine interfaces
  • Image and Signal processing
  • Machine learning and pattern recognition
  • Robotics
  • Human-robot interaction
  • Handwriting recognition
  • Speech recognition

Selected Talks

Data Sets

  • "Signatures", Data sets of signatures acquired with the camera-based interface for handwriting described above.

Author: Mario E. Munich <mariomu AT ieee DOT org>

Date: 2010-03-23 00:42:27 PDT

HTML generated by org-mode 6.34c in emacs 23