| The sixth national population census’ data in 2012 shows that with the degree of growing aging population, the number of old people aged 60 and above in China was over 194 million. Besides, due to nerve and musculoskeletal system diseases, natural disasters and accidents, the number of patients suffering permanent disability is increasing year by year. The growing of aging people and the disabled results in the increasement of permanently disabled patient population, and its proportion of China’s total population is also on the rise. With the degree of the increasing permanently disabled patient population getting more serious in China, the society and their families are facing a heavy burden and a severe test to maintain their daily life. A lot of foucus from the government and the public have been made to solve his social problem and help those permanently disabled patients to maintain their daily life and communicate with the outside world independently.Based on the physical properties of head electromagnetic signals, the dissertation takes advantage of magnetic signal with the spatial distribution characteristic and bioelectric signal with the lumped parameter characteristic, including the marked magnetic signal for tongue motion tracking, bioelectric signal from eye movement, and electroencephalography (EEG) or magnetoencephalography (MEG) derived from neural activities in brain, and carries out research on control commands extraction for human-machine interaction system (HMIS). To solve the key problem that how to extract accurate and multi-model control commands in real-time, the dissertation carries on theory researches, conductes validations based on simulation and experiment, and then applies the extracted commands for practical HMIS:to meet the real-time requirement in command extraction of magnetic marked tongue motion interface (TMI), we propose a real-time orientation-invariant magnetic localization method based on closed-form models; to meet the multi-model requirement in command extraction of hybrid brain-machine interface (BMI), we propose a multi-model feature signal classification method based on the time/frequency characteristics of bioelectric signals from single-channel dry-electrode; to meet the accurate requirement in command extraction of MEG based HMI, we propose a method for magnetic source localization based on reconstructed magnetic field.The remainder of the dissertation offers the following:In Chapter 1, the background and significance of the research were introduced, along with the development trend and current research situations of control command extraction of HMIS based on TMI, EMI and BMI which serve permanently disabled patients of different levels. Then the research contents of this dissertation were proposed.In Chapter 2, the theory research on feature signal extraction based on electromagnetic signal reconstruction was proposed. Firstly, in magnetic localization for TMI, a closed-form inverse model based on orientation-invariant magnetic source was proposed, and a magnetic source upright device was designed, along with unique solution determination and sensor system calibration method, which realizes real-time magnetic localization with high accuracy. Secondly, in lumped parameter feature extraction of hybrid BMI, a multi-model control commands extraction method based on time/frequency characteristics of bioelectric signals from single-channel dry-electrode was proposed, where eye-blink-derived control commands were extracted according to the amplitude and time-interval characteristics in time domain and concentration-derived control commands were extracted according to frequency characteristics of EEG Alpha rhythm in frequency domain. Finally, in active neural source localization of MEG based BMI, a localization method, which used reconstructed magnetic field in the curl-free space by solving a Laplace’s equation with measured sparse boundary conditions (BCs), was introduced. BC interpolation based on polynomial fitting and reconstruction selection criteria based on the localization improvement ratio and gradient of the reconstructed magnetic fields were proposed, along with nonlinear magnetic localization method based on least-square technique.In Chapter 3, simulations and experiments were carried out to validate the control commands extraction method based on electromagnetic signal reconstruction. Firstly, by using a PM marker, a PM-upright device to maintain a constant magnetic dipole moment was proposed, and experiments on different movement traces were conducted to validate the magnetic localization method based on closed-form inverse model. Secondly, bioelectric signal sampling system based on a single-channel dry-electrode was developed, and human experiments were carried out on the able-bodied and disabled subjects, where taking account of individual differences, pre-calibration paradigm for feature signal extraction was designed. Finally, in simulation validation of magnetic source localization based on reconstructed magnetic field, the common magnetic dipole model was employed and results verified the "smooth" effect brought by reconstruction and the improvement on localization accuracy.In Chapter 4, applications of control commands extraction method based on electromagnetic field reconstruction in HMIS were introduced. Firstly, using the quantized eye-blink commands provided by the single-channel dry-electrode system, experiments on an electric wheelchair are conducted. Secondly, using multi-model control commands derived from hybrid bioelectric signals measured by a single-channel dry-electrode, experiments on a 6-DOF robot arm were conducted, and a two-level control strategy based on the multi-model control commands was developed to promote the automatic level and operation efficiency of the system. Then, active neural source localization in MEG-based HMI was proposed, where the magnetic fields outside the head were reconstructed using dense BC interpolated from sparse measurements, and improved localization computed from reconstructed fields with high-fidelity was provided for MEG-based HMI.In Chapter 5, the chief work and innovations of this dissertation were summarized, and the further research subjects were proposed.
|
PDF Full Text Request