PhD thesis

Contributions in systems based on Virtual Reality techniques used in biomedical field

Dan Marius Dobrea

    On 6th may 2005 it was the public presentation of the doctoral thesis. The Ph.D. thesis was validated by the National Council for the Attestation of Academic Degrees, Diplomas and Certificates and I received the Magna Cum Laude grade.

Doctoral thesis resume    

    The Ph.D. thesis proposes a new bioinstrumental system capable to determine the subject's fatigue state as a result of human-system interaction (e.g.human - computer interaction).
    In the classification process the bioinstrumental system used three psychological signals acquired without any contact with the subject. These signals were: hand tremor signal, respirator signal and the subject body movements. The thesis presents: the system of sensors, acquisition techniques, the preprocessing methods used to artifact elimination, the extracted features from the psychological signals, the methodology used to reveal the tired state, the methods used to analyze the fatigue state, the classification systems, the obtained results, the analyses of these results and, in the end, the conclusions that can be used to further improve the system's performances.
    After on an overview of the HCI fields, in the second chapter the static and the dynamic characteristic of a new class of resonant sensors are extracted and presented. This sensor was used to record the hand tremor signal and the respiratory activity. These extracted characteristics prove the transducer performances and, more, the sensor capability to acquire both physiological signals (named above) without any quality loses and components distortions. This mode of behave is very important mainly because even small components can convey important information that can be use by the pattern recognition system. In order to extract the sensor's dynamic characteristic a custom system was build around of the TMS320F240 DSP system. This system is capable to generate a reference mechanical movement accordingly with the user requirements. Moreover, in this system it is incorporated a new developed RNSIC converter (Rectifier With Near-Sinusoidal Current) able to improve the waveform of the current drawn from the power source.
    In 3rd chapter a new noncontact Virtual Joystick system is presented starting with the system's main concepts, the development stages of different parts of the system and the device analysis. The Virtual Joystick uses three transducers similars with the transducer presented in the previously chapter. The Virtual Joystick system is presented starting with: the transducers command and control units, the external interfacing units (the heart of this unit is the TMS320F240 DSP), the PC software application (it presents the hand position from the 3D input space and manages the hand tremor signal acquisition), the first dedicated fuzzy system used to compensate sensor characteristic and the noise suppression and, finally, the second fuzzy system used to modeled the hand position and to determine correctly the hand position. In the end, the acquired tremor signal is analyzed in order to estimate the system's ability to obtain a "real" tremor signal. These results outline once again the Virtual Joystick performances and capabilities.
    Chapter 4 is dedicated to the presentation and the analysis of the biological process known under the name of "tremor". After a short overview of the general elements and a description of the tremor characteristics the existing components of the tremor signal are presented (mechanical components, components generated by the neuro-muscular feedback, central components, etc.). In the following part of the chapter the methodology used to evidence the fatigue state is presented. The problems of subjects' selection and the standard rules in the fields, the protocol and data acquiring methodologies, methods of data preprocessing, features extraction, fatigue state analyses, classes analyses and the classificatory selection are presented, all these in order to correctly differentiate the fatigue state. Moreover, this chapter gives an idea of a large diversity of information that is embedded in the tremor signal and highlights the possibility to recognize the fatigue state using for this only the tremor signal.
    In chapter 5 there are presented, properly explained, applied and analyzed several methods of signal processing in order to investigate the results obtained in the previously chapter. The size of data set is studied and the assumptions relevance regarding the tired state is also analyzed. The central nervous influence on the tremor signal is deeply investigated using several different methods. All three method used (frequency analyses, coherence analyses and neural network analyzes) converge to the same result: the physiological tremor signal has a central nervous system origin.
    The Chapter 6 is dedicated to the presentation of the contribution regarding the design, construction and testing of a new system capable to acquire the respiratory activity without any kind of contact using the same sensor presented and analyzed in the Chapter 2. The respiratory signal acquired with this system is contaminated with two types of artifacts: with slow variation and by fast artifacts (these slow and fast variations are compared with the respiratory behavioral speed). The slow artifacts are eliminated based on a custom hardware and software implemented adaptive system. This method shows superior performances compared with a classical filter and with software implemented adaptive system. In the next section the fast artifacts are eliminated based on a Blind Source Separation method. This method proves to be more robust and accurate in artifact removing, in the frame of nonstationary artifacts compared with neural networks systems. Using these methods of artifact removing the new proposed system is capable to obtain a respiratory signal unaffected by two large classes of artifacts.
    The seven chapters entitled "A noncontact laser system for body language acquisition and interpretation of subject movements" present a new innovative system of human computer interaction. After an introduction the system working mode and implementation are presented. This system was implemented in two different versions. In the first implementation the system's core was a PC and in the second embodiment a TMS 320C6414 DSP was used. This system offers for the first time in the field of HCI systems the possibility to use a new kind of information regarding the subject's emotional state, unexploited yet on this domain, namely, the emotional state of the subject expressed through his/her body language.
    In the last chapter the conclusions and the contribution of this Ph.D. thesis are highlighted and the future directions of studies (having the obtained results as a starting point) are presented.