As neural networks are trained to perform tasks of increasing complexity, their size increases, which presents several challenges in their deployment on devices with limited resources. To cope with this, a recently proposed approach hinges on substituting the classical Multiply-and-ACcumulate (MAC) neurons in the hidden layers with other neurons called Multiply-And-Max/min (MAM) whose selective behavior helps identify important interconnections, thus allowing aggressive pruning of the others. Hybrid MAM&MAC structures promise a 10x or even 100x reduction in their memory footprint compared to what can be obtained by pruning MAC-only structures. However, a cornerstone of maintaining this promise is the assumption that MAC&MAM architectures have the same expressive power as MAC-only ones. To concretize such a cornerstone, we take here a step in the theoretical characterization of the capabilities of mixed MAM&MAC networks. We prove, with two theorems, that two hidden MAM layers followed by a MAC neuron with possibly a normalization stage is a universal approximator.

Accelerated MRI acquisition is widely adopted and basically consists in undersampling the current slice at the cost of a quality degradation. What samples to skip is determined by an encoder, while the quality loss is partially compensated by the use of a decoder. The hypothesis behind accelerated MRI acquisition is that to higher acceleration factors always correspond lower reconstruction qualities with an undersampling pattern that is usually fixed at design time, neglecting adaptability on the slice acquired at inference time. This paper proposes a novel accelerated MRI acquisition method that enables single-slice adaptation by dividing the acquisition into incremental batches and estimating the reconstruction quality at the end of each batch. The acquisition terminates as soon as the target quality is reached. We demonstrate the efficacy of our novel method using a state-of-the-art neural model capable of jointly optimizing the encoder and decoder. To estimate the current quality of the slice we reconstruct and propose a neural quality predictor. We demonstrate the advantages of our novel acquisition method compared to classic acquisition for two different datasets and for both line-constrained and unconstrained Cartesian sampling strategies (theoretically implementable via 2D and 3D imaging respectively). • A novel incremental MRI acquisition method with adaptive undersampling. • Utilizes neural self-assessment to adjust MRI acquisition dynamically. • Introduction of target reconstruction quality, tunable before acquisition. • Reaching target quality more efficiently than classic acquisition. • Validated on FastMRI and IXI datasets: showcasing robust performance across settings.

The growing interest in the Internet of Things (IoT) and mobile artificial intelligence applications is pushing the investigation on deep neural networks (DNNs) that can operate at the edge using low-resources/energy devices. To obtain such a goal, several pruning techniques have been proposed in the literature. They aim to reduce the number of interconnections-and consequently the size, and the corresponding computing and storage requirements-of DNNs that traditionally rely on classic multiply-and-accumulate (MAC) neurons. In this work, we propose a novel neuron structure based on a multiply-and-max/min (MAM) map-reduce paradigm, and we show that by exploiting this new paradigm it is possible to build naturally and aggressively prunable DNN layers, with a negligible loss in performance. This novel structure allows a greater interconnection sparsity when compared to classic MAC-based DNN layers. Moreover, most of the already existing state-of-the-art pruning techniques can be used with MAM layers with little to no changes. To test the pruning performance of MAM, we employ different models-AlexNet, VGG-16 and the more recent ViT-B/16-and different computer vision datasets-CIFAR-10, CIFAR-100, and ImageNet-1K. Multiple pruning approaches are applied, ranging from single-shot methods to training-dependent and iterative techniques. As a notable example, we test MAM on the ViT-B/16 model fine-tuned on the ImageNet-1K task and apply one-shot gradient-based pruning. We remove interconnections until the model experiences a 6% decrease in accuracy. While the selected MAC-based layers need at least 38.2% remaining interconnections, MAM-based layers achieve the same accuracy with only 0.1%.

Anomalies are rare events, and this affects the design flow of detectors that monitor systems that behave normally most of the time but whose failure may have serious consequences. This limitation is particularly evident in the detector performance evaluation: it requires an abundance of normal and anomalous data but realistically faces a scarcity of the latter. To address this, in this article, we develop a framework comprising a set of abstract anomalies modeling the effects real-world failures and disturbances have on sensor readings. In addition, we devise synthetic generation procedures for these anomalies. Given a dataset of normal tracks from the actual application, one may apply such procedures to produce anomalous-like time series for a comprehensive detector assessment. We show that this framework can anticipate the detector performing best with real-world anomalies in the context of human and structural health monitoring, also highlighting that, in these cases, the best detector is not the most complex.

Large monitoring systems produce data that is often compressed to be transmitted over the network. For latency or security reasons, compressed data may be processed at the edge, i.e., along the path from sensors to the cloud, for some purposes such as anomaly detection. However, the performance of a detector distinguishing between normal and anomalous behavior may be affected by the loss of information due to compression. We here analyze how lossy compression affects the performance of a generic anomaly detector. This relationship is formalized in terms of information-theoretic quantities. Within such a framework we leverage a Gaussian assumption to derive analytical results regarding the importance of white noise as a representative of both the average and asymptotic anomalies. Moreover, in an anomaly-agnostic scenario, we also show the existence of a level of compression for which an anomaly is undetectable though compression is not completely destructive. Numerical evidence confirms that the proposed information-theoretic quantities anticipate the performance of practical compressors and detectors in the case of Gaussian and non-Gaussian signals allowing an assessment of the tradeoff between compression and detection.

Anomaly Detection is a challenging task due to the limited knowledge about possible anomalies. This issue can be tackled by modeling anomalies through domain expertise or collecting sufficient anomalous data. However, some domains, such as monitoring systems, require detectors that are capable of detecting any potential alteration in the observed phenomenon. Hereby we propose a tool to generate anomalies as a statistical deviation from the characterization of the signal representing the normal behavior. Two families of deviation models are presented, and the effectiveness of the tool is proven using well-known unsupervised detectors. The effects of a possible intermediate data compression stage on the detection capabilities are also considered.

Due to its intrinsic sparsity both in time and space, event-based data is optimally suited for edge-computing applications that require low power and low latency. Time varying signals encoded with this data representation are best processed with Spiking Neural Networks (SNN). In particular, recurrent SNNs (RSNNs) can solve temporal tasks using a relatively low number of parameters, and therefore support their hardware implementation in resource-constrained computing architectures. These premises propel the need of exploring the properties of these kinds of structures on low-power processing systems to test their limits both in terms of computational accuracy and resource consumption, without having to resort to full-custom implementations. In this work, we implemented an RSNN model on a low-end, resource-constrained ARM-Cortex-M4-based Micro Controller Unit (MCU). We trained it on a down-sampled version of the N-MNIST event-based dataset for digit recognition as an example to assess its performance in the inference phase. With an accuracy of 97.2%, the implementation has an average energy consumption as low as$4.1\ \mu\mathrm{J}$and a worst-case computational time of$150.4\ \mu\mathrm{s}$per time-step with an operating frequency of 180 MHz, so the deployment of RSNNs on MCU devices is a feasible option for small image vision real-time tasks.

Subspace analysis (SA) is a widely used technique for coping with high-dimensional data and is becoming a fundamental step in the early treatment of many signal-processing tasks. However, traditional SA often requires a large amount of memory and computational resources, as it is equivalent to eigenspace determination. To address this issue, specializedstreamingalgorithms have been developed, allowing SA to be run on low-power devices, such as sensors or edge devices. Here, we present a classification and a comparison of these methods by providing a consistent description and highlighting their features and similarities. We also evaluate their performance in the task of subspace identification with a focus on computational complexity and memory footprint for different signal dimensions. Additionally, we test the implementation of these algorithms on common hardware platforms typically employed for sensors andedgedevices.

The search for energy efficient circuital implementations of neural networks has led to the exploration of phase-change memory (PCM) devices as their synaptic element, with the advantage of compact size and compatibility with CMOS fabrication technologies. In this work, we describe a methodology that, starting from measurements performed on a set of real PCM devices, enables the training of a neural network. The core of the procedure is the creation of a computational model, sufficiently general to include the effect of unwanted non-idealities, such as the voltage dependence of the conductances and the presence of surrounding circuitry. Results show that, depending on the task at hand, a different level of accuracy is required in the PCM model applied at train-time to match the performance of a traditional, reference network. Moreover, the trained networks are robust to the perturbation of the weight values, up to 10% standard deviation, with performance losses within 3.5% for the accuracy in the classification task being considered and an increase of the regression RMS error by 0.014 in a second task. The considered perturbation is compatible with the performance of state-of-the-art PCM programming techniques.

The increasing need for small and low-power Deep Neural Networks (DNNs) for edge computing applications involves the investigation of new architectures that allow good performance on low-resources/mobile devices. To this aim, many different structures have been proposed in the literature, mainly targeting the reduction in the costs introduced by the Multiply and Accumulate (MAC) primitive. In this work, a DNN layer based on the novel Sum and Max (SAM) paradigm is proposed. It does not require either the use of multiplications or the insertion of complex non-linear operations. Furthermore, it is especially prone to aggressive pruning, thus needing a very low number of parameters to work. The layer is tested on a simple classification task and its cost is compared with a classic DNN layer with equivalent accuracy based on the MAC primitive, in order to assess the reduction of resources that the use of this new structure could introduce.

Several recent works have shown the advantages of using phase-change memory (PCM) in developing brain-inspired computing approaches. In particular, PCM cells have been applied to the direct computation of matrix-vector multiplications in the analog domain. However, the intrinsic nonlinearity of these cells with respect to the applied voltage is detrimental. In this paper we consider a PCM array as the encoder in a Compressed Sensing (CS) acquisition system, and investigate the effect of the non-linearity of the cells. We introduce a CS decoding strategy that is able to compensate for PCM nonlinearities by means of an iterative approach. At each step, the current signal estimate is used to approximate the average behaviour of the PCM cells used in the encoder. Monte Carlo simulations relying on a PCM model extracted from an STMicrolectronics 90 nm BCD chip validate the performance of the algorithm with various degrees of nonlinearities, showing up to 35 dB increase in median performance as compared to standard decoding procedures.

Principal component analysis (PCA) is a well-established approach commonly used for dimensionality reduction. However, its computational cost and memory requirements hamper the adoption of PCA in heavily resource-constrained embedded platforms. Streaming approaches have been proposed that may enable embedded implementations of the PCA. Among them, the history PCA (HPCA) algorithm stands out for its robustness to the variability in parameters and accuracy. This article presents a parallel and memory-efficient implementation of HPCA in a structural health monitoring (SHM) application based on a heterogeneous network with sensor nodes measuring three-axial accelerations and gateways collecting measurements from several nodes and sending them to the cloud storage and analytic facility. In the targeted application, standard PCA reaches 15x compression factor with an average reconstruction signal-to-noise ratio of 16 dB and a negligible impact on the accuracy in the tracking of structural modal frequencies. By embedding HPCA on our SHM network gateways, we achieve the same compression factor as standard PCA, with more than 1000x reduction in data memory footprint for running the algorithm. Furthermore, we parallelize HPCA on the gateway, and we achieve a speedup of 7.1x (on 8 cores). Finally, we explore a fixed-point HPCA implementation on sensors (network end nodes), that maximally distributes compression workload, minimizes required communication bandwidth, and maintains the same quality of reconstruction as HPCA in floating point, with a compression factor of 10x.

The main challenge in the implementation of long-lasting vibration monitoring systems is to tackle the constantly evolving complexity of modern “mesoscale” structures. Thus, the design of energy-aware solutions is promoted for the joint optimization of data sampling rates, onboard storage requirements, and communication data payloads. In this context, the present work explores the feasibility of the rakeness-based compressed sensing (Rak-CS) approach to tune the sensing mechanism on the second-order statistics of measured data. In particular, a novel model-assisted variant (MRak-CS) is proposed, which is built on a synthetic derivation of the spectral profile of the structure by pivoting on numerical priors. Moreover, a signal-adapted sparsity basis relying on the wavelet packet transform operator is conceived, which aims at maximizing the signal sparsity while allowing for a precise time-frequency localization. The adopted solutions were tested with experiments performed on a sensorized pinned-pinned steel beam. Results prove that the rakeness-based compression strategies are superior to conventional eigenvalue approaches and to standard CS methods. The achieved compression ratio is equal to seven and the quality of the reconstructed structural parameters is preserved even in presence of defective configurations.

In a recent paper, a through-the-barrier communication technique suitable for isolated resonant converters has been proposed. The approach is capable of sending data bidirectionally at high speed (one bit for each converter clock period) without the need of any additional isolating device other than the transformer necessary for the power transfer, and has been demonstrated by means of a proof-of-concept low-frequency prototype. In this paper we review that work under the assumption of increasing the operating frequency by using a coreless transformer presenting low losses, but also a low coupling factor k. This allows to increase the efficiency of the converter to a very high value (92% in the proposed design working at 6.78 MHz), but the communication speed has to be reduced (one bit every four clock cycles).

Approaches for reconstructing signals encoded with Compressed Sensing (CS) techniques, and based on Deep Neural Networks (DNNs) are receiving increasing interest in the literature. In a recent work, a new DNN-based method named Trained CS with Support Oracle (TCSSO) is introduced, relying the signal reconstruction on the two separate tasks of support identification and measurements decoding. The aim of this paper is to improve the TCSSO framework by considering actual implementations using a finite-precision hardware. Solutions with low memory footprint and low computation requirements by employing fixed-point notation and by reducing the number of bits employed are considered. Results using synthetic electrocardiogram (ECG) signals as a case study show that this approach, even when used in a constrained-resources scenario, still outperform current state-of-art CS approaches.

The amount of data generated by distributed monitoring systems that can be exploited for anomaly detection, along with real time, bandwidth, and scalability requirements leads to the abandonment of centralized approaches in favor of processing closer to where data are generated. This increases the interest in algorithms coping with the limited computational resources of gateways or sensor nodes. We here propose two dual and lightweight methods for anomaly detection based on generalized spectral analysis. We monitor the signal energy laying along with the principal and anti-principal signal subspaces, and call for an anomaly when such energy changes significantly with respect to normal conditions. A streaming approach for the online estimation of the needed subspaces is also proposed. The methods are tested by applying them to synthetic data and real-world sensor readings. The synthetic setting is used for design space exploration and highlights the tradeoff between accuracy and computational cost. The real-world example deals with structural health monitoring and shows how, despite the extremely low computations costs, our methods are able to detect permanent and transient anomalies that would classically be detected by full spectral analysis.

Compressed sensing can be used to yield both compression and a limited form of security to the readings of sensors. This can be most useful when designing the low-resources sensor nodes that are the backbone of IoT applications. Here, we propose to use chaining of subsequent plaintexts to improve the robustness of CS-based encryption against ciphertext-only attacks, known-plaintext attacks and man-in-the-middle attacks.

This paper presents the design of an ultra-low energy, rakeness-based compressed sensing (CS) system that utilizes time-mode (TM) signal processing (TMSP). To realize TM CS operation, the presented implementation makes use of monostable multivibrator based analog-to-time converters, fixed-width pulse generators, basic digital gates and an asynchronous time-to-digital converter. The TM CS system was designed in a standard 0.18 μm IC process and operates from a supply voltage of 0.6V. The system is designed to accommodate data from 128 individual sensors and outputs 9-bit digital words with an average reconstruction SNR of 35.31 dB, a compression ratio of 3.2, with an energy dissipation per channel per measurement vector of 0.621 pJ at a rate of 2.23 k measurement vectors per second.

A recent paper improved the state of the art for resonant class-E dc/dc converters by relaying their design on the solution of an associated non-linear dimensionless mathematical system. We show in this paper that when the associated nonlinear system can be solved, the solution is not always unique. By considering a simple case study we are able to compute two different solutions leading to two different designs. In the first one the main spectral component of voltage and current waveforms is located around the first clock harmonic, and around the second harmonic in the other solution. Interestingly, the latter design leads to a reduction either in the size of inductors/transformers or in the oscillation frequency (a 2.46× factor), and also a non-negligible improvement in the converter efficiency (from 74.1% to 75.8%).

Atrial electrogram (AEG) acquired with a high spatio-temporal resolution is a promising approach for early detection of atrial fibrillation. Due to the high data rate, transmission of AEG signals requires considerable energy, making its adoption a challenge for low-power wireless devices. In this paper, we investigate the feasibility of using compressed sensing (CS) for the acquisition of AEGs while reducing redundant data without losing information. We apply two CS approaches, standard CS and rakeness-based CS (rak-CS) on real medical recordings. We find that the AEGs are compressible in time, and, more interestingly, in the spatial domain. The performance of rak-CS is better than standard CS, especially at higher compression ratios (CR), both during sinus rhythm (SR) and atrial fibrillation (AF). More specifically, the difference in the achieved average reconstruction signal-to-noise (ARSNR) in rak-CS and standard CS, for CR = 4.26, in the time domain is 7.7 dB and 2.6 dB for AF and SR, respectively. Multi-channel data is modeled as a multiple-measurement-vector problem and a suitable mixed norm is used to exploit the group structure of the signals in the spatial domain to obtain improved reconstruction performance over l1norm minimization. Using the mixed-norm recovery approach, for CR = 4.26, the difference in achieved ARSNR performance between rak-CS and standard CS is 5 dB and 2 dB for AF and SR, respectively.

Chaining, i.e., the mode of operation in which each message is encrypted considering a digital summary of previous ones, is here applied to block-cipher stages based on compressed sensing. We show that this simple and parsimonious technique may significantly harden the resulting system with respect to common threats such that ciphertext-only, known-plaintext, and man-in-the-middle attacks. Non-negligible robustness comes at the price of not more than a 2% of energy overhead with respect to the pure compression stage which represents a 24× reduction with respect to straightforward implementation of a traditional cryptography primitive like Advanced Encryption Standard.

Compressed Sensing (CS) has recently emerged as an effective way to simultaneously acquire, compress and possibly encrypt incoming signals in low-resource sensing devices. We here show that CS can be suitably exploited to add disturbance rejection properties, similar to those which are classically obtained by means of suitably designed and deployed signal conditioning stages. This may render such additional stages unnecessary and therefore substantial decrease both system complexity and energy requirements. An example dealing with electrocardiographic signals is developed in which the classical base-line and power-line disturbances are almost entirely rejected with no need of ad-hoc filters.

The incoming Internet of Things revolution requires the adoption of innovative paradigms for the design of low-power ubiquitous sensor nodes. This can be achieved by exploiting Compressed Sensing (CS), that is a recently introduced approach capable of simultaneously sampling and compressing an input signal with a limited amount of resources. While the underlying basic theory is well developed, in recent years we have seen a flourishing of CS techniques capable of exploiting some additional priors on the input signal to improve performance. In this paper, we propose a survey and a comparison of the most promising ones. We use a classification mechanism based on which prior is used and which processing block is modified with respect to the standard CS.

In common distributed sensing scenarios, a number of local wireless sensor networks perform sets of acquisitions that must be sent to a central collector which may be far from the measurement fields. Hence, readings from individual nodes may reach their destination by exploiting both local and long-range transmission capabilities. The compressed sensing (CS) paradigm may help finding a convenient mix of the two options, especially if it follows the rakeness-based design flow that has been recently introduced. CS is exploited by identifying local hubs that aggregate many sensor readings in a smaller number of quantities that are then transmitted to the central collector. We here show that, depending on the relative cost of local versus long-range transmission, carefully administering the choice of the hubs, the breadth of the neighborhood from which they collect readings, as well as the coefficients with which those readings a linearly aggregated, one may significantly reduce the energy needed to sample the field. Simulations indicate that savings may be over 50% for values of the parameters modeling nowadays local and long-range transmission technologies.

Compressed sensing has been proposed to both yield low-cost compression and low-cost encryption. This can be very useful in the design of sensor nodes with a limited resource budget whose acquisition must be kept as private as possible. We here analyze the susceptibility of compressed sensing stages that are optimized to maximize compression performance by rakeness-based design to ciphertext-only and known-plaintext attacks. A tradeoff between compression and security is highlighted. Notwithstanding such a tradeoff, rakeness-based compressed sensing exhibits a noteworthy robustness to classical attacks.

Compressed Sensing (CS) is an effective way to sample a signal at a sub-Nyquist rate, i.e., by using a number of measurements smaller than the number of samples required when using the standard Nyquist approach. Measurements are obtained as linear projections of input signals along random sensing vectors. CS has been often regarded as a democratic method, in the sense that each measurement contributes to signal reconstruction with a similar amount of information. In this paper, by combining empirical observations with results from recent papers, we propose a different point of view, and show that CS is an oligarchic approach where performance is basically set by the measurements with the highest energy. This allows us to propose a new CS-based approach that bases the reconstruction on the maximum-energy measurements only and improves the compression performance with respect to classical approaches.

Wireless sensor nodes capable of acquiring and transmitting biosignals are increasingly important to address future needs in healthcare monitoring. One of the main issues in designing these systems is the unavoidable energy constraint due to the limited battery lifetime, which strictly limits the amount of data that may be transmitted. Compressed Sensing (CS) is an emerging technique for introducing low-power, real-time compression of the acquired signals before transmission. The recently developed rakeness approach is capable of further increasing CS performance. In this paper we apply the rakeness-CS technique to enhance compression capabilities for electroencephalographic (EEG) signals, and particularly for Evoked Potentials (EP), which are recordings of the neural activity evoked by the presentation of a stimulus. Simulation results demonstrate that EPs are correctly reconstructed using rakeness-CS with a compression factor of 16. Additionally, some interesting denoising capabilities are identified: the high-frequency noise components are rejected and the 60 Hz power line noise is decreased by more than 20dB with respect to the state-of-the-art filtering when rakeness-CS techniques are applied to the EEG data stream.

Compressed sensing, further to its ability of reducing resources spent in signal acquisition, may be seen as an implicit private-key encryption scheme. The level of achievable secrecy has been analyzed in the most classical settings, when the sensing matrix is made of independent and identically distributed entries. Yet, it is known that substantially improved acquisition can be achieved by tuning the statistics of such a matrix. The effect of such an optimization on the robustness with respect to classical cryptographic attacks is analyzed here.

Compressed Sensing (CS) has recently emerged as an interesting and effective way to sample an input signal and at the same time compress it (i.e., reduce the number of measurements for the correct signal reconstruction with respect to the standard Nyquist approach). We show here that CS can be used also to exploit some operations typically performed by the preceding signal conditioning stage (sometimes, by a post-processing stage). In detail, we show that CS can be used to filter environmental disturbances exactly like a notch filter. Furthermore, this solution presents advantages in terms of input signal distortion with respect to the classical notch filter approach. An example on electrocardiographic signal is presented as case study.

We report a low cost mobile EEG system for characterizing cortical auditory responses. The system is built using commercial off-the-shelf components and each unit costs less than $200. It measures seven EEG channels plus one audio channel (envelope only), and communicates the data to external devices via Bluetooth. A novel implementation was pursued in order to support local signal compression using compressed sensing. At the same time, it provides a low cost solution that is useful for recording cortical auditory responses and extracting clinically relevant features of the waveform. This system has been designed with the eventual goal of long term monitoring of the brain activity of schizophrenic patients outside a clinical setting, in order to better understand auditory hallucinations and manage their ongoing treatment. In this preliminary study we obtained simultaneous audio and cortical recordings of evoked auditory responses from normal healthy subjects wearing the EEG for several hours in duration. We report evoked auditory responses for 2 Hz and 40 Hz click trains. We also report alpha wave responses, demonstrating stable and high quality recordings over a five hour period.

The sensitivity of recovery algorithms with respect to a perfect knowledge of the encoding matrix is a general issue in many application scenarios in which compressed sensing is an option to acquire or encode natural signals. Quantifying this sensitivity in order to predict the result of signal recovery is therefore valuable when no a priori information can be exploited, e.g., when the encoding matrix is randomly perturbed without any exploitable structure. We tackle this aspect by means of a simplified model for the signal recovery problem, which enables the derivation of an average performance estimate that depends only on the interaction between the sensing and perturbation matrices. The effectiveness of the resulting heuristic is demonstrated by numerical exploration of signal recovery under three simple perturbation matrix models. Finally, we show how this estimate matches very well the degradation experienced by non-perfectly informed decoders in applications of compressed sensing to protecting the acquired information content in ECG tracks and sensitive images.

Power-Line Communication (PLC) systems represent a very interesting opportunity for introducing low-cost communication capabilities over already existing power-line wires. In this paper we introduce a PLC technique that can be applied to systems with a DC power bus that employ a switching power converter as main power supply unit. The proposed technique is extremely simple to be implemented, and requires only minor modifications on the main switching converter. As a proof of this, we are capable to implement the proposed PLC in a system composed by commercial DC-DC converter boards without any circuital modification to the boards themselves. Measurements on this test system show the capability to communicate up to about 80 kbit/s with a bit error rate so low as 10−5.

When transmission or storage costs are an issue, lossy data compression enters the processing chain of resource-constrained sensor nodes. However, their limited computational power imposes the use of encoding strategies based on a small number of digital computations. In this case study, we propose the use of an embodiment of compressed sensing as a lossy digital signal compression, whose encoding stage only requires a number of fixed-point accumulations that is linear in the dimension of the encoded signal. We support this design with some evidence that for the task of compressing ECG signals, the simplicity of this scheme is well-balanced by its achieved code rates when its performances are compared against those of conventional signal compression techniques.

Despite the linearity of its encoding, compressed sensing (CS) may be used to provide a limited form of data protection when random encoding matrices are used to produce sets of low-dimensional measurements (ciphertexts). In this paper, we quantify by theoretical means the resistance of the least complex form of this kind of encoding against known-plaintext attacks. For both standard CS with antipodal random matrices and recent multiclass encryption schemes based on it, we show how the number of candidate encoding matrices that match a typical plaintext-ciphertext pair is so large that the search for the true encoding matrix inconclusive. Such results on the practical ineffectiveness of known-plaintext attacks underlie the fact that even closely related signal recovery under encoding matrix uncertainty is doomed to fail. Practical attacks are then exemplified by applying CS with antipodal random matrices as a multiclass encryption scheme to signals such as images and electrocardiographic tracks, showing that the extracted information on the true encoding matrix from a plaintext-ciphertext pair leads to no significant signal recovery quality increase. This theoretical and empirical evidence clarifies that, although not perfectly secure, both standard CS and multiclass encryption schemes feature a noteworthy level of security against known-plaintext attacks, therefore increasing its appeal as a negligible-cost encryption method for resource-limited sensing applications.

In this work we combing two novelty in the area of Analog Information Converter based on Compressed Sensing. A new architecture, the Spread Spectrum Random Modulation PreIntegration and a new design flow, the rakeness based design of a Compressed Sensing system. We demonstrate that combining these approaches produces a strong reduction of the internal chipping frequency in the sensing coupled with a high compression ratio with respect to standard Analog to Digital Converter.

Classical design of Analog-to-Information converters based on Compressive Sensing uses random projection matrices made of independent and identically distributed entries. Leveraging on previous work, we define a complete and extremely simple design flow that quantifies the statistical dependencies in projection matrices allowing the exploitation of non-uniformities in the distribution of the energy of the input signal. The energy-driven reconstruction concept and the effect of this design technique are justified and demonstrated by simulations reporting conspicuous savings in the number of measurements needed for signal reconstruction that approach 50%.

Though compressive sensing hinges on extracting linear measurements from the signals to acquire, actual implementations introduce nonlinearities whose effect can be far from negligible. We here address the problem of saturation in the circuit blocks needed by a Random Modulation Pre-Integration architecture. To allow a fair a comparison with previous analysis, we rely on a model capturing the essentials of saturations in actual implementations while being able to reproduce more abstract settings considered in the literature. Based on this, we analyze some methods already proposed to cope with simplified saturation mechanisms, briefly discussing their underlying principles. Finally, we introduce a novel approach that takes into account the more realistic model and, at the cost of an almost negligible hardware overhead, is extremely effective in countering saturation effects.

Analog to Information conversion is a new paradigm in signal digitalization. In this framework, compressed sensing theory allows to reconstruct sparse signal from a limited number of measures. In this work, we will assume that the signal is not only sparse but also localized in a given domain, so that its energy is concentrated in a subspace. We will present a formal and quantitative discussion to explain how localization of sparse signals can be exploited to improve the quality of the reconstructed signal.

Calibration or model parameter estimation from measured data is an ubiquitous problem in engineering. In systems biology this problem turns out to be particularly challenging due to very short data-records, low signal-to-noise ratio of data acquisition, large intrinsic process noise and limited measurement access to only a few, of sometimes several hundreds, state variables. We review state-of-the-art model calibration techniques and also discuss their relation to the general reverse-engineering problem in systems biology. For biomolecular circuits involving low-copy-number molecules we adopt a Markov process setup and discuss a calibration approach based on suitable metrics between probability measures and propose the metrics computation for the multivariate case. In particular, we use Kantorovich's distance and devise an algorithm, for the case when FACS (fluorescence-activated cell sorting) measurements are given. We discuss a case study involving FACS data for the high-osmolarity glycerol (HOG) pathway in budding yeast.

This paper presents a way to reduce the effect of narrowband interference (due to either an intentional jamming or the effect of traditional non-spread-spectrum transmission) in ultra-wideband (UWB) communication systems based on asynchronous direct-sequence code-division multiple access (DS-CDMA). To reach this goal, we derive a closed-form expression for the bit error probability in additive white Gaussian noise (AWGN) channel, where both multiple access and narrowband interference are the main cause of nonideality. By leveraging on this, we develop a new approach based on spectrum shaping of the spreading sequences waveforms which allows to significantly improve performance in many applicative scenarios.