Research On Gas Sensor Signal Sampling System Based On CS Theory

With the rapid development of modern electronic&information technology, all kinds of information and data are needed, the Nyquist theorem, as a basic guiding, has brought great pressure and challenges to the signals processing and hardware. However, to recover the original signal, the Nyquist theorem is sufficient but not necessary. In recent years, a new data acquisition theory named compressive sensing has been proposed, the prevalence of CS has broken through the restriction of the Nyquist theorem, which demand the sampling frequency is at least twice the minimum frequency of the signal, also it can effectively solve the current problems.According to the sparsity and the statistical property of the original analog signal, specially gas sensor signals, in the paper, we propose a sampling implement based on compressive sensing theory; and in the back-end, we focus on the reconstruction algorithm, an improved algorithm is proposed, which performance is better than the origin algorithm, specially to the original sparse analog signals. In the final part, we studied the possible structure of the proposed sampling implement. The main work is as follows:1. Study on compressed sensing theory. Focus on the sparse representation、observation matrix and reconstruction algorithm aspects of the compressive sensing theory, we especially research the mismatch of sparse representation basis, and the impact on the reconstruction algorithm; and then, we propose an improved reconstruction algorithm for the compressed sampling signal. Different from the classical reconstruction algorithm, the algorithm proposed deploy the improved orthodox match pursuit algorithm, which firstly reconstruct the sparse coefficients in the frequency domain, and then obtain the original signals through inverse transform. Since the proposed algorithm avoids the problem of sparse representation basis mismatch, the reconstruction accuracy is improved. For the observation matrix, we give the design principles and constrains detail, and discuss the physical implementation form.2. Study on the random sampling method based on compressive sensing. In this paper, we first analyzes the basic structure and mathematical model of the analog to information convector based on the CS theory; basing on these, we research the parallel analog information convector and segmented parallel compressive sampling model further, the compressive matrix and measurement matrix of each model is given. Considering both the achievable of hardware and the spectral coefficient characteristics of the signal, we proposed a prototype hardware using Hadamard sequence, which based on the compressive sensing theory and easy to achieve. We analysis the prototype in the frequency domain, it is proving that the random sequence can reduce the correlation between the Hadamard sequence and sparse representation, finally, we derive the mathematical representation of the compressive sensing matrices.3. Study of the physical implementation of random sampling. Aiming at the former prototype hardware we proposed, we study the physics of this model, and analysis the property of the random sequence based on the Hadamard matrix, furthermore, we preliminary research its physical implementation.