LAUSR

Dear editor, As preliminary implementation of compressed sensing (CS), the analog-to-information converter (AIC) plays an important role in reducing the sampling rate of analog signals via PN sequence [1, 2]. However, it suffers severe reconstruction errors (REs) when acquiring noisy signals prior to measurement in cases with low signal-noise ratio (SNR) [3]. To damp the RE, the adaptive weighted norm constraint is first introduced in [4], which enables the weighted l1 norm constrained model obtain an accurate approximation of the l0 norm model. Distilled sensing is used to control the noise folding by designing adaptive measurements to detect and locate weak signals in the additive white Gaussian noise (AWGN) environment [5]. Further, two new-types of decoding procedures are proposed by combining l1-minimization with either a regularized selective least p-powers or an iterative hard thresholding [6]. Meanwhile, a data pre-processing operation, such as the adaptive selective compressive sampling (ASCS) in [7], is preliminarily considered in the recovery algorithm. These relative studies on the RE are mainly focused on the sparse signal and its noise; however, it is the nonlinear reconstruction matrix that plays the dominant role in the RE [8]. In this study, the single-channel sampling and multi-channel reconstruction (SCS-MCR) AIC scheme is proposed to damp the RE completely with low hardware consumption, as shown in Figure 1(a). In the proposed scheme, the nonlinear reconstruction matrix is approximate to the linear diagonal matrix by the incoherent accumulation in the digital domain. The proposed model. The proposed SCS-MCR AIC scheme is depicted in Figure 1(a). First, the multiple mixing sequences are superimposed together on the original analog signal x(t). Further, the mixed analog signal is sampled by only one analog-to-digital converter (ADC) after the lowpass filter. Finally, in the digital domain, the recovery signals are accumulated together after they are reconstructed in each channel according to each mixing sequence. The mathematical process is realized as follows. As shown in Figure 1(a), the original signal x ∈