Reconstruction And Nonlinear-Correction For Iot-Sensor-Signal Under Sparse Sampling

With the continuous development of technology and applications, the IOT (Internet of the things) brings a profound change to people’s lives and work. There are a lot of sensors needed in IOT, which collect data and send these data to the central processing system. Then the central processing system studies the data, thus ensuring IOT runs exactly. However,there are commonly some non-linearity in sensors of temperature, humidity, current and other IOT sensors.Traditional IOT sensors’ non-linearity correction method requires high-rate signal acquisition and signal conversion in data acquisition boards,also owns the ability to carry out large-scale data analysis during non-linearity correction. So the topic of using the compressed sensing theory to overcome these two difficulties during IOT sensors’non-linearity correction presented in this paper is very meaningful and innovative.For reconstruction algorithm, this paper does deep research on OMP(Orthogonal Matching Pursuit) and Dual Affine Scaling Interior Point Method;through MATLAB simulation program, we do more comprehensive comparasion on these two reconstruction strategies in regard to Reconstruct Signal-Noise Ratio, Means Squared Error,Percentage of Energy Recovery.The result shows that, OMP algorithm performs more stable than Dual Affine Scaling Interior Point Method.But with respect to the former, Dual Affine Scaling Interior Point Method needs less sparse-sampling-data to reconstruct. The minimum data length Dual Affine Scaling Interior Point Method needed is only26.27%of the original data sampled based on Shannon-Nyquist Theorem.For sparse sampling strategy, this paper proposes random equivalent time sampling strategy, and does detailed theoretical explanations. Simulation and analysis of this strategy by MATLAB shows the impact of this stratagy on required data length of compressed sensing reconstruction algorithm.This paper proposes that there are two sampling method exsited:"Full-Line" sampling method and Adaptive-Sampling Methods and shows the trends of Reconstruct Signal-Noise Ratio, Means Squared Error.Percentage of Energy Recovery toward sampling step in "Full-Line" samples.Also shows the desired data length reconstruction needed is11.11%of the original data length with the optimization of Adaptive-Sampling Methods.Meanwhile, this paper analyzes the Random-Equivalent-Time-Sampling-Process implemented in STC12C5A60S2and timing to transmit data through the serial port.In this system, current, temperature, humidity and other IOT sensor data are transmitted to the host computer by PL2303; then do recontruction by theses data through MATLAB. This system has a goog effect in reconstruction.For IOT sensors nonlinearity correction, this paper studies the stratagy based on Support Vector Machine to do nonlinearity correction. And build non-linear sensor calibration models and algorithms based on Support Vector Machine algorithm to do the non-linear compensation toward Random-Equivalent-Time-Sampling-Process.For sensor data collecting in IOT system, IOT sensor-nonlinear-calibration and related applications, the research in this paper owns significance.