Digital Signal Processing 81 (2018) 1–7
Contents lists available at ScienceDirect
Digital Signal Processing
www.elsevier.com/locate/dsp
Compressive sampling for spectrally sparse signal recovery via one-bit
random demodulator
Han-Fei Zhou
a
, Lei Huang
a,∗
Jian Li
b
a
Institute of Multi-dimensional Signal Processing, College of Information Engineering, Shenzhen University, Shenzhen, Guangdong, 518000, China
b
Department of Electrical and Computer Engineering, University of Florida, Gainesville, FL 32611-6130, United States of America
a r t i c l e i n f o a b s t r a c t
Article history:
Available
online 18 May 2018
Keywords:
Spectrally
sparse signals
One-bit
compressive sampling
Random
demodulator (RD)
Binary
iterative hard thresholding (BIHT)
The one-bit compressive sampling (CS) framework aims at alleviating the quantization burden on analog-
to-digital
converters by quantizing each sample to one bit, i.e., capturing just the signs of samples.
Motivated by one-bit CS theory, this paper addresses a new type of data acquisition system to recover
spectrally sparse signals. This system is composed of a random demodulator and a one-bit quantizer.
The former yields the signal compressed samples while the latter records the sign of each sample.
With the observation sign data, the signal is eventually recovered by using the binary iterative hard
thresholding algorithm. Through numerical experiments, we demonstrate that our scheme is high-
efficient
for spectrally sparse signal recovery in the situations of low signal-to-noise ratio, stringent bit
budget and weak sparsity.
© 2018 Elsevier Inc. All rights reserved.
1. Introduction
Shannon sampling theorem provides one of the foundations in
modern signal processing. For a continuous real-valued time signal
f (t) whose highest frequency is not larger than W /2hertz, the
sampling theorem suggests that we sample the signal uniformly at
a rate equal to W hertz. At a sampling rate of W , the values of the
signal at intermediate points are determined completely by
f (t) =
n∈Z
f (n/W )sinc(Wt −n), (1)
where Z denotes the set of integer and sinc(x) = sin(x)/x.
In
the absence of extra information, Nyquist-rate sampling can
completely recover f (t). However, this well-known approach be-
comes
impractical when the bandwidth W becomes very large
because it is challenging for sampling hardware to operate at such
a high rate. Even though recent developments in analog-to-digital
converter (ADC) technologies have boosted sampling speeds, they
cannot meet the requirements in many real-world applications,
such as ultra-wideband communications and cognitive radar etc.,
in which the signal bandwidth might be larger than 10 GHz. On
the other hand, the high power consumption is another important
restriction, which prohibits the ADC from many wide-band appli-
cations
[1].
*
Corresponding author.
E-mail
address: lhuang@szu.edu.cn (L. Huang).
Fortunately, in many implementations, the signals are spectrally
sparse, that is, they can be approximated as harmonic tones. In-
spired
by compressive sampling (CS) theory [2,3], Tropp et al. [4]
have
designed a new type of sampling system, called random de-
modulator
(RD), which exploits the spectral sparsity. It is shown
in [4] that the RD requires just O (K log(W /K )) samples per sec-
ond
to reliably reconstruct the signal, where K is the number of
significant frequency tones.
In
practice, the samples must be quantized, i.e., each sample is
mapped from a continuous value with infinite precision to a dis-
crete
value with some finite precision. In [5,6], X. Gu et al. have
analyzed the recovery performance of quantized compressed sens-
ing
(QCS). The most extreme form of quantization is reducing the
signal to one-bit for each sample, which may be accomplished
by repeatedly comparing the signal to some reference level, and
recording whether the signal is above or below it [7,8]. In [9],
Jacques et al. have bridged the one-bit and high-resolution quan-
tized
CS theories.
One-bit
CS theory extends the CS framework to the extreme
quantization, and it shows that some signal parameters can be
recovered with high accuracy from samples of which each is quan-
tized
to just one bit [10–20]. The one-bit CS theory is of two
important advantages in practical implementations. First, simple
one-bit hardware quantizers consist of only a comparator and can
operate at high speeds. Thus, we can reduce the sampling com-
plexity
by reducing the bit depth, rather than decreasing the num-
ber
of measurements [21]. Second, because the one-bit encoding is
https://doi.org/10.1016/j.dsp.2018.04.014
1051-2004/
© 2018 Elsevier Inc. All rights reserved.