Hardware-Algorithms Co-Design and Implementation of an Analog-to-Information Converter for Biosignals Based on Compressed Sensing

Abstract

We report the design and implementation of an Analog-to-Information Converter (AIC) based on Compressed Sensing (CS). The system is realized in a CMOS 180 nm technology and targets the acquisition of bio-signals with Nyquist frequency up to 100 kHz. To maximize performance and reduce hardware complexity, we co-design hardware together with acquisition and reconstruction algorithms. The resulting AIC outperforms previously proposed solutions mainly thanks to two key features. First, we adopt a novel method to deal with saturations in the computation of CS measurements. This allows no loss in performance even when 60% of measurements saturate. Second, the system is able to adapt itself to the energy distribution of the input by exploiting the so-called rakeness to maximize the amount of information contained in the measurements. With this approach, the 16 measurement channels integrated into a single device are expected to allow the acquisition and the correct reconstruction of most biomedical signals. As a case study, measurements on real electrocardiograms (ECGs) and electromyograms (EMGs) show signals that these can be reconstructed without any noticeable degradation with a compression rate, respectively, of 8 and 10.

Non-Uniform Wavelet Sampling for RF Analog-to-Information Conversion

“…We start by introducing the basics of CS and then review the most prominent A2I converter architectures for RF signal acquisition, namely non-uniform sampling (NUS) [15], [16], [26]- [29], variable rate sub-Nyquist sampling [8], [30]- [32], and random modulation [33], [34], which includes the modulated wideband converter and Xampling [11], [35]- [37]. For each of these architectures, we briefly discuss the pros and cons from a RF spectrum sensing and hardware design standpoint.…”

“…Existing architectures first multiply the analog input signal by a pseudo-random sequence, integrate the product over a finite time window, and sample the integration result. The random-modulation preintegrator (RMPI) [5], [34] and its single branch counterpart, the random demodulator (RD) [18], [33], are the most basic instances of this idea. However, modulating the signal with a (pseudo-)random sequence is only suitable for very specific signal classes, such as signals that are well-represented by a union of sub-spaces [37].…”

Energy Analysis of Decoders for Rakeness-Based Compressed Sensing of ECG Signals

“…Intriguingly, A can also be made only of antipodal symbols, i.e., A ∈ {−1, +1} m×n . This constraint is of paramount importance as it allows hardware-friendly architectures, where expensive and cumbersome full multipliers are not required anymore, and represents a key point in the design of effective and parsimonious CS stages for biomedical sensing nodes [13]. For this reason, in this manuscript we assume that the projection matrix A is antipodal.…”

“…Several works have appeared in the recent literature proposing low-power CS encoders with particular interest in biomedical applications [9]. Commonly, Electroencephalographic (EEG) [10], [11] or Electrocardiographic (ECG) signals [12], [13] are considered. However, a very limited number of works can be found proposing energy considerations on the decoder.…”

Communication Channel Analysis and Real Time Compressed Sensing for High Density Neural Recording Devices

“…Analog implementation performs multiplication and accumulation using analog switches and integrators [28] [29] [30]. Digital implementation moves the multiplication and accumulation step to the digital domain after A/D conversion [8] [16] [17].…”