Modeling, Analysis Of Surface Bioelectric Signal And Its Application In Human Computer Interaction

The surface bioelectrical signal is the refection of the self-consciousness of alife entity. Currently two kinds of the surface bioelectrical signals, named Elec-troencephalography (EEG) and Electromyography (EMG), are typically used forHuman-Computer Interaction (HCI). However, the understanding of the bioelec-trical signal is still not clear. Some characteristics of the signal itself and theintroduction of various types of noise in the acquisition process make the surfacebioelectrical signals based HCI is still in the laboratory research stage. Thereare still some problems for the practical application. The research in this thesisconsiders the application of HCI with two sources of the surface bioelectric sig-nals, which are EEG and EMG. Motor imagery and Steady State Visual EvokedPotential (SSVEP) are considered as the research objects in Brain ComputerInterface (BCI). The main works are as follows,In the issue of the motor imagery based BCI, we frstly study the general-ization ability of the Common Spatial Pattern (CSP) algorithm. By applyingthe linear mixture model, simulation data is used to compare the impact of var-ious factors on the generalization ability of the algorithm. Secondly, we extendCSP algorithm to less channel condition, and make the algorithm not only havethe ability to run in the less channel condition but also be able to deal withfrequency-domain information. Finally, the realization of the online platformvalidates the efectiveness.In the issue of the SSVEP based BCI, we frstly discuss the infuence of theduty cycle on SSVEP through experiments. Furthermore, based on a series ofexperiments and assumptions, we simulate the generation of SSVEP by coupledneural mass model. Finally, we also use the online platform to validate the efectiveness of the system.For EMG signal, considering the muscle distribution characteristics andcrosstalk problem of EMG signals in the progress of propagation, we introducethe commonly used spatial flter methods in BCI into EMG signal processing.Using Common Spatio-Spectral Pattern algorithm, the classifcation result is im-proved by the use of both spatial and spectral information. Finally, we also tryto identify the handwriting movements used in daily life by EMG. Dynamic timewarping algorithm are used for the overall recognition of handwritten signal ofthe EMG.