Proc. SPIE. 8759, Eighth International Symposium on Precision Engineering Measurement and Instrumentation
KEYWORDS: Signal to noise ratio, Digital signal processing, Linear filtering, Demodulation, Signal processing, Analog electronics, Electronic filtering, Detection theory, Filtering (signal processing), Compressed sensing
Random Demodulator (RD) is a novel signal acquisition scheme based on compressed sensing (CS). It is able to acquire signals continuous in time and sparse in frequency at a sampling rate far below the Nyquist frequency. In the RD architecture, low-pass filter plays a role of anti-aliasing and is also the main part of the measurement matrix which should be characterized accurately. In this paper, we analyze the impact of the low-pass filter’s non-idealities. If the filter parameters deviate from their ideal values, there will be mismatch between the ideal impulse response used in the measurement matrix and the practical, which will degrade the reconstruction performance. The results show that with the increase of the degree of mismatch, the quality of signal reconstruction, measured by signal to noise ratio (SNR), declines. The results suggest accurate calibration is needed in practical use.
Underdetermined blind mixing model recovery (UBMMR) is one of the most important steps in separating
underdetermined blind sources, which has a direct effect on the recovery accuracy of source signals. A new blind mixing
model recovery algorithm is proposed, under the assumption that the sources are sparse. The mixture data observed are
first allocated to several clusters using the partitional clustering algorithm based on differential evolution (DE). The
cluster centers are amended through Hough transformation to recover the mixing model. The peak clustering problem in
Hough transformation is successfully avoided at the same time. Experimental results show that the proposed algorithm
has advantages of high robustness and accuracy compared with conventional algorithms.