Hardware implementation of memristor-based artificial neural networks.

Artificial Intelligence (AI) is currently experiencing a bloom driven by deep learning (DL) techniques, which rely on networks of connected simple computing units operating in parallel. The low communication bandwidth between memory and processing units in conventional von Neumann machines does not support the requirements of emerging applications that rely extensively on large sets of data. More recent computing paradigms, such as high parallelization and near-memory computing, help alleviate the data communication bottleneck to some extent, but paradigm- shifting concepts are required. Memristors, a novel beyond-complementary metal-oxide-semiconductor (CMOS) technology, are a promising choice for memory devices due to their unique intrinsic device-level properties, enabling both storing and computing with a small, massively-parallel footprint at low power. Theoretically, this directly translates to a major boost in energy efficiency and computational throughput, but various practical challenges remain. In this work we review the latest efforts for achieving hardware-based memristive artificial neural networks (ANNs), describing with detail the working principia of each block and the different design alternatives with their own advantages and disadvantages, as well as the tools required for accurate estimation of performance metrics. Ultimately, we aim to provide a comprehensive protocol of the materials and methods involved in memristive neural networks to those aiming to start working in this field and the experts looking for a holistic approach.

Spread-Spectrum Modulated Multi-Channel Biosignal Acquisition Using a Shared Analog CMOS Front-End.

The key challenges in designing a multi-channel biosignal acquisition system for an ambulatory or invasive medical application with a high channel count are reducing the power consumption, area consumption and the outgoing wire count. This article proposes a spread-spectrum modulated biosignal acquisition system using a shared amplifier and an analog-to-digital converter (ADC). We propose a design method to optimize a recording system for a given application based on the required SNR performance, number of inputs, and area. The proposed method is tested and validated on real pre-recorded atrial electrograms and achieves an average percentage root-mean-square difference (PRD) performance of 2.65% and 3.02% for sinus rhythm (SR) and atrial fibrillation (AF), respectively by using pseudo-random binary-sequence (PRBS) codes with a code-length of 511, for 16 inputs. We implement a 4-input spread-spectrum analog front-end in a 0.18 µm CMOS process to demonstrate the proposed approach. The analog front-end consists of a shared amplifier, a 2nd order Σ∆ ADC sampled at 7.8 MHz, used for digitization, and an on-chip 7-bit PRBS generator. It achieves a number-of-inputs to outgoing-wire ratio of 4:1 while consuming 23 µA/input including biasing from a 1.8 V power supply and 0.067 mm2 in area.

From Chaos to Pseudorandomness: A Case Study on the 2-D Coupled Map Lattice.

Applying the chaos theory for secure digital communications is promising and it is well acknowledged that in such applications the underlying chaotic systems should be carefully chosen. However, the requirements imposed on the chaotic systems are usually heuristic, without theoretic guarantee for the resultant communication scheme. Among all the primitives for secure communications, it is well accepted that (pseudo) random numbers are most essential. Taking the well-studied 2-D coupled map lattice (2D CML) as an example, this article performs a theoretical study toward pseudorandom number generation with the 2D CML. In so doing, an analytical expression of the Lyapunov exponent (LE) spectrum of the 2D CML is first derived. Using the LEs, one can configure system parameters to ensure the 2D CML only exhibits complex dynamic behavior, and then collect pseudorandom numbers from the system orbits. Moreover, based on the observation that least significant bit distributes more evenly in the (pseudo) random distribution, an extraction algorithm E is developed with the property that when applied to the orbits of the 2D CML, it can squeeze uniform bits. In implementation, if fixed-point arithmetic is used in binary format with a precision of z bits after the radix point, E can ensure that the deviation of the squeezed bits is bounded by 2-z . Further simulation results demonstrate that the new method not only guides the 2D CML model to exhibit complex dynamic behavior but also generates uniformly distributed independent bits with good efficiency. In particular, the squeezed pseudorandom bits can pass both NIST 800-22 and TestU01 test suites in various settings. This study thereby provides a theoretical basis for effectively applying the 2D CML to secure communications.

Deep Neural Oracles for Short-Window Optimized Compressed Sensing of Biosignals.

The recovery of sparse signals given their linear mapping on lower-dimensional spaces can be partitioned into a support estimation phase and a coefficient estimation phase. We propose to estimate the support with an oracle based on a deep neural network trained jointly with the linear mapping at the encoder. The divination of the oracle is then used to estimate the coefficients by pseudo-inversion. This architecture allows the definition of an encoding-decoding scheme with state-of-the-art recovery capabilities when applied to biological signals such as ECG and EEG, thus allowing extremely low-complex encoders. As an additional feature, oracle-based recovery is able to self-assess, by indicating with remarkable accuracy chunks of signals that may have been reconstructed with a non-satisfactory quality. This self-assessment capability is unique in the CS literature and paves the way for further improvements depending on the requirements of the specific application. As an example, our scheme is able to satisfyingly compress by a factor of 2.67 an ECG or EEG signal with a complexity equivalent to only 24 signed sums per processed sample.

On the Security of a Class of Diffusion Mechanisms for Image Encryption.

The need for fast and strong image cryptosystems motivates researchers to develop new techniques to apply traditional cryptographic primitives in order to exploit the intrinsic features of digital images. One of the most popular and mature technique is the use of complex dynamic phenomena, including chaotic orbits and quantum walks, to generate the required key stream. In this paper, under the assumption of plaintext attacks we investigate the security of a classic diffusion mechanism (and of its variants) used as the core cryptographic primitive in some image cryptosystems based on the aforementioned complex dynamic phenomena. We have theoretically found that regardless of the key schedule process, the data complexity for recovering each element of the equivalent secret key from these diffusion mechanisms is only (1). The proposed analysis is validated by means of numerical examples. Some additional cryptographic applications of this paper are also discussed.

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

In recent years, compressed sensing (CS) has proved to be effective in lowering the power consumption of sensing nodes in biomedical signal processing devices. This is due to the fact the CS is capable of reducing the amount of data to be transmitted to ensure correct reconstruction of the acquired waveforms. Rakeness-based CS has been introduced to further reduce the amount of transmitted data by exploiting the uneven distribution to the sensed signal energy. Yet, so far no thorough analysis exists on the impact of its adoption on CS decoder performance. The latter point is of great importance, since body-area sensor network architectures may include intermediate gateway nodes that receive and reconstruct signals to provide local services before relaying data to a remote server. In this paper, we fill this gap by showing that rakeness-based design also improves reconstruction performance. We quantify these findings in the case of ECG signals and when a variety of reconstruction algorithms are used either in a low-power microcontroller or a heterogeneous mobile computing platform.

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

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.