Electrical Impedance Tomography Technology And Its Application In Human-Computer Interaction

Electrical Impedance Tomography(EIT)is a non-invasive functional imaging technology that can reconstruct the internal electrical impedance distribution of the tar-get object from the boundary measurement data.Compared with the common imag-ing technology,electrical impedance imaging technology has the advantages of non-destructive,non-invasive,non-radiation,low cost and real-time characteristics.There-fore,it is widely used in biomedical imaging,geological studies and industrial nonde-structive testing.In recent years,the rapid development of human-computer interaction technol-ogy has brought attention to the application of EIT technology in tactile sensing and gesture recognition.On the one hand,the most widely used flexible sensors are the flex-ible array sensors consisting of series of sensing units.However,the array sensors are complex with tiny communication units that make manufactures complicated and ex-pensive.Additionally,in order to transmit data from large-scale sensor array,a large number of wires is usually arranged inside the sensor.The distribution of wires will not only cause electromagnetic noise,but also reduce the flexibility and stretchability,thus limiting the applications in practice.While the EIT based tactile sensor can avoid these shortcomings.On the other hand,the non-invasive and fast response characteristics of EIT can make it suitable for measuring the bioelectrical impedance of the human wrist,providing a new and simpler gesture recognition method for HCI.However,EIT is a serious non-linear,ill-conditioned inverse problem,causing the poor spatial resolution and contrast of the image obtained by traditional reconstruction methods.In addition,the application of EIT technology in human-computer interaction still has some ques-tions such as low accuracy and real-time performance.Therefore,we have conducted the following research around EIT and its application in human-computer interaction.1,we studied the mathematical principles of the EIT algorithm,and proposed a multi-frame block constrained Sparse Bayesian learning algorithm to improve the resolution and contrast of the reconstructed image.The Sparse Bayesian learning method can automatically explore the block sparsity and correlation information of the impedance distribution in the measured object.From a statistical point of view,we per-form mathematical transformations on the inverse problem solving process,and obtain a maximum posterior probability estimation problem under certain constraints.Com-bined with the block structure a prior information,the impedance distribution image can be effectively reconstructed.Simulation and experiment results demonstrate that the proposed method can improve the image resolution and remove the artifacts and noises interference.2,we studied the application of EIT technology in tactile sensing.We designed a portable tactile sensor system,and proposed two contact detection algorithms.By dividing the sensing part of the tactile sensor into multiple sub-regions,we adapted ma-chine learning methods to transform the inverse problem into a classification problem.The supervised learning method was adopted for training and learning,which can re-alize fast and accurate position detection.The second method is to optimize the sparse Bayesian learning to improve the real-time performance of imaging.Both experiments and simulations verify that the EIT tactile sensor can achieve fast and accurate tactile sensing.3,we studied the application of EIT technology in gesture recognition and pro-posed two different gesture recognition systems.There is a difference of impedance data utilization when the collected boundary measurement data were used to image re-construction and gesture recognition.Therefore,an EIT drive pattern based on model explanation and feature selection is proposed,which can greatly reduce the complexity of the data acquisition system while ensuring the accuracy of recognition and real-time recognition.On this basis,we proposed an EIT based two-electrode frequency-scan ges-ture recognition system.By exploring the characteristics between impedance data and frequency changes under different gestures,a supervised learning method is adopted to achieve accurate gesture recognition.In the dexterous hand control experiment,the accuracy and real-time performance of the system were verified.