Hardware Implementation Complexity Research Articles

High-speed PON downstream transmission based on pre-configured KK scheme with CD pre-compensation and direct detection

In this paper, the pre-configured Kramers–Kronig (Pre-KK) scheme combined with the chromatic dispersion pre-compensation (CDPC) has been proposed and studied to reduce the complexity of DSP and hardware implementation at the optical network unit (ONU) side for passive optical network (PON), avoiding the use of KK receiver. For further analysis and verification, 100 Gb/s/λ PON downstream transmissions with 4-level pulse amplitude modulation (PAM-4) signal based on the Pre-KK scheme and KK receiver have been implemented and compared on a unified optical system platform. The results show that the Pre-KK requires 1-bit resolution lower than KK receiver for ADC, and achieves a power budget of 29 dB after 20 km fiber transmission at the strong FEC threshold of 1 × 10−2. By increasing the launch power, the maximum power budget of 33.2 dB can be achieved due to the fiber nonlinearity. The effectiveness of CDPC in Pre-KK scheme has also been investigated and demonstrated over 20 and 50 km transmission. Besides, the computational complexities of two schemes are analyzed and discussed in detail. Hence, the Pre-KK scheme joint with CDPC can be a promising solution for the high-speed PON downstream transmission.

Optimized Continuous Wavelet Transform Algorithm Architecture and Implementation on FPGA for Motion Artifact Rejection in Radar-Based Vital Signs Monitoring

The continuous wavelet transform (CWT) has been used in radar-based vital signs detection to identify and to remove the motion artifacts from the received radar signals. Since the CWT algorithm is computationally heavy, the processing of this algorithm typically results in long processing time and complex hardware implementation. The algorithm in its standard form typically uses software processing tools and is unable to support high-performance data processing. The aim of this research is to design an optimized CWT algorithm architecture to implement it on Field Programmable Gate Array (FPGA) in order to identify the unwanted movement introduced in the retrieved vital signs signals. The optimization approaches in the new implementation structure are based on utilizing the frequency domain processing, optimizing the required number of operations and implementing parallel processing of independent operations. Our design achieves significant processing speed and logic utilization optimization. It is found that processing the algorithm using our proposed hardware architecture is 48 times faster than processing it usingMATLAB. It also achieves an improvement of 58% in speed performance compared to alternative solutions reported in literature. Moreover, efficient resources utilization is achieved and reported. This advanced performance of the proposed design is due to consciously implementing comprehensive approaches of multiple optimization techniques that results in multidimensional improvements. As a result, our achieved design is suitable for utilization in high-performance data processing applications.

Cooperative Detection Method for Distracted and Fatigued Driving Behaviors With Readily Embedded System Implementation

To mitigate risks caused by distracted and fatigued driving, timely and accurate detection of these behaviors is significant for safety concern. Among the state of the art of detection methods based on in-car image analysis, the interaction of multiple distracted and fatigued driving behaviors could make the certain behavior-targeted detection ineffective and unreliable, while the complicated neural-network-based detection methods could be poor in interpretability and feasibility of hardware implementation. This article proposes a novel cooperative detection method for distracted and fatigued driving behaviors, with overall consideration of method performance, operation complexity, and practical hardware implementation. The vital points for detection of hand-held calling, looking left/right continuously, yawning, eye closure, and the cooperative relation among different behaviors are investigated. Experiment cases on the established dataset which involves indoor driving simulation images and actual cockpit driving scenario images of different drivers under various illuminations and background settings are conducted. On the established Di_Fa_C_Tes dataset, the proposed method achieves above 98% detection precision and 37 frames/s processing speed for the tested behaviors by support vector machine (SVM) on a PC platform. Additionally, the article provides a hardware evaluation of the proposed method on i.MX 8QuadMax platform and reaches above 96.8% precision and 16 frames/s processing speed. The experimental results indicate the effectiveness of the proposed method in accurate and fast detection of distracted and fatigued driving behaviors and the promising embedded system application.

Robust Hybrid Precoding Design for Securing Millimeter-Wave IoT Networks Under Secrecy Outage Constraint

Hybrid precoding architecture, as a cost-effective approach for millimeter-wave (mmWave) communications, can achieve an excellent tradeoff between spectrum efficiency and hardware implementation complexity. However, the design of a robust hybrid precoding for improving the physical layer security (PLS), which is insensitive to the uncertainty of eavesdropper's channel state information (CSI), has not been well studied. This article for the first time designs a probabilistically robust hybrid precoding scheme for securing broadcast communications in Internet of Things (IoT) with eavesdropper's imperfect CSI. Specifically, considering the Gaussian CSI error model, we maximize the minimum secrecy rate of multiple IoT devices (IoDs) by jointly designing analog and digital precoders under the constraints in terms of secrecy outage probability and per IoD's information rate. The optimization problem is challenging due to the coupling of the analog and digital precoders, and the secrecy outage constraint. To handle these challenges, we first employ a conservative probability inequality to transform the secrecy outage probability constraint into a deterministic one. Then, by employing the penalty dual decomposition (PDD) method, we develop a novel iterative algorithm to convert the resultant nonconvex problem into a sequence of convex problems, which can guarantee the convergence to its Karush-Kuhn-Tucker (KKT) solution. Simulation results show that the proposed algorithm can achieve significant secrecy performance gains compared with the benchmark algorithm.

Defining of the Minimum Wavelet Filter Coefficients Bit-Width for 3D Medical Imaging

Medical imaging uses a variety of modalities to provide visual information about a patient. Various methods are used to process this data. Many of them are based on discrete wavelet transform (DWT). Its use will allow effective denoising and compression of 2D and 3D images. This paper proposes a new approach to linear time-invariant wavelet filtering using quantized filter coefficients when using which the computational errors have different signs and allow to partially compensate each other as a result of which the processed image is of high quality. The analysis of the quantization noise of the direct multilevel DWT filter coefficients is carried out. The derived formulas demonstrate the relationship between the quantization accuracy of these coefficients and the processing quality of digital 3D images. The derived formulas for calculating the minimum accuracy of the wavelet filter coefficients representation in the computing devices memory allow minimizing the effect of quantization noise on the result of 3D images processing. Modelling of 3D medical tomographic images DWT processing showed that a decrease in the ratio of the average voxel brightness to the maximum allowable value with increasing color depth of images leads to faster achievement of high quality compared to the results of theoretical analysis with an increase in the value of the scaling degree of the wavelet filter coefficients. The obtained theoretical and practical results open up the possibility for reducing the computational complexity of software and hardware implementation of wavelet processing of 3D medical visual data on modern microelectronic devices (field-programmable gate arrays, application-specific integrated circuits, etc.).

Versatile Configuration and Control Framework for Real-Time Data Acquisition Systems

Modern physics experiments often utilize FPGA-based systems for real-time data acquisition. Integrated analog electronics demand for complex calibration routines. Furthermore, versatile configuration and control of the whole system is a key requirement. Beside low-level register interface to the FPGA, also access to I$^2$C and SPI buses is often needed to configure the complete system. Calibration through an FPGA is inflexible and yields a complex hardware implementation. On the contrary, calibration through a remote system is possible but considerably slower due to repetitive network accesses. By using SoC-FPGA solutions with a microprocessor, more sophisticated configuration and calibration solutions, as well as standard remote access protocols, can be efficiently integrated in software. Based on Xilinx Zynq US+ SoC-FPGAs, we implemented a versatile control framework. This software framework offers a convenient access to the hardware and a flexible abstraction via remote-procedure calls (RPCs). Based on the open source RPC library gRPC, functionality with low-latent control flow, complex algorithms, data conversions and processing, as well as configuration via external buses can be provided to a client via Ethernet. Furthermore, client interfaces for various programming languages can be generated automatically which eases collaboration among different working groups and integration into existing software. This contribution presents the framework as well as benchmarks regarding latency and data throughput.

Digital interpolation-based SFO compensation in OCT precoding-enabled NHS-OFDM transmission systems

The non-Hermitian symmetric optical orthogonal frequency division multiplexing (NHS-OFDM) system exhibits lower hardware implementation complexity and power consumption as maintaining the same bandwidth granularity, compared to the conventional Hermitian symmetric optical OFDM (HS-OFDM) system. Also, NHS-OFDM can obtain a similar peak-to-average power ratio as well as bit error rate (BER) performance. It is expected to be a new candidate for the next-generation passive optical network. In the asynchronous NHS-OFDM system, the sampling clocks for the analog to digital converter in the receiver and the digital to analog converter in the transmitter generated from different sources and their frequencies always have a certain degree of deviation, i.e., sampling frequency offset (SFO). It may significantly deteriorate the receiver performance. Besides, the imperfect frequency response of the optical/electrical devices is another factor for the degraded BER performance. In this work, we experimentally investigate the digital interpolation (DI)-based SFO compensation in a short-reach NHS-OFDM transmission system. The orthogonal circulant matrix transform (OCT) precoding technique is employed to equalize the signal-to-noise ratio over data-carrying subcarriers and then improve the BER performance. The experimental results indicate that, with the help of the 4-th order DI, the SFO up to 200 ppm can be well compensated without power penalties in comparison with the SFO-free case. Furthermore, the receiver sensitivity of the NHS-OFDM enabled by OCT precoding can be improved by more than 2 dB at the BER of 1e-3 after 20 km standard single-mode fiber transmission.

A new multi-scroll Chua’s circuit with composite hyperbolic tangent-cubic nonlinearity: Complex dynamics, Hardware implementation and Image encryption application

In this paper, a new method for generating multi-scroll chaotic attractors via constructing a compound hyperbolic tangent-cubic nonlinear function in canonical Chua’s circuit is presented. The basic dynamic characteristics of the system are analyzed, including equilibrium points, bifurcation diagrams, Lyapunov exponents, phase portraits, time-domain diagrams and attractive basins. What is surprising is that the proposed multi-scroll Chua’s circuit also exhibits rich dynamic behaviors like coexisting multiple attractors, transient period, intermittent chaos and offset boosting. In addition, we put forward the application of the system in chaotic image encryption, and analyzed some security performance evaluation indexes to show that the new Chua’s chaotic cipher system has high security and reliable encryption performance. Finally, the hardware design and experiments of the chaotic digital circuits and image encryption are carried out. Both numerical simulation and FPGA experimental results verify the feasibility and usability of the proposed new multi-scroll Chua’s system.

Memristor Based Binary Convolutional Neural Network Architecture With Configurable Neurons.

The memristor-based convolutional neural network (CNN) gives full play to the advantages of memristive devices, such as low power consumption, high integration density, and strong network recognition capability. Consequently, it is very suitable for building a wearable embedded application system and has broad application prospects in image classification, speech recognition, and other fields. However, limited by the manufacturing process of memristive devices, high-precision weight devices are currently difficult to be applied in large-scale. In the same time, high-precision neuron activation function also further increases the complexity of network hardware implementation. In response to this, this paper proposes a configurable full-binary convolutional neural network (CFB-CNN) architecture, whose inputs, weights, and neurons are all binary values. The neurons are proportionally configured to two modes for different non-ideal situations. The architecture performance is verified based on the MNIST data set, and the influence of device yield and resistance fluctuations under different neuron configurations on network performance is also analyzed. The results show that the recognition accuracy of the 2-layer network is about 98.2%. When the yield rate is about 64% and the hidden neuron mode is configured as −1 and +1, namely ±1 MD, the CFB-CNN architecture achieves about 91.28% recognition accuracy. Whereas the resistance variation is about 26% and the hidden neuron mode configuration is 0 and 1, namely 01 MD, the CFB-CNN architecture gains about 93.43% recognition accuracy. Furthermore, memristors have been demonstrated as one of the most promising devices in neuromorphic computing for its synaptic plasticity. Therefore, the CFB-CNN architecture based on memristor is SNN-compatible, which is verified using the number of pulses to encode pixel values in this paper.

Simple blind estimation of sampling frequency offset for real-time asynchronous DDO-OFDM system

Sampling frequency offset (SFO) will result in phase rotation, inter-channel interference, and inter-symbol interference, which can seriously degrade the transmission performance of the asynchronous direct-detection optical OFDM (DDO-OFDM). In this paper, we study a simple and overhead-free SFO blind estimation technique in a DDO-OFDM system. The key parameters in this technique are optimized by theoretical analysis, numerical simulation, and experiments. A simplified method of the SFO blind estimation is also proposed to further reduce the hardware implementation complexity. The accuracy and stability of the SFO estimation are experimentally investigated in the real-time asynchronous DDO-OFDM transmission system. The results indicate that the SFO estimation accuracy can be within ±1.5 ppm and shows high stability even at an SFO of up to 200 ppm after 20-km standard single-mode fiber transmission.

Surface Recognition via Force-Sensory Walking-Pattern Classification for Biped Robot

Real-time surface recognition has become a critical factor for ensuring safe walking of intelligent biped robots in a complex human living environment. This work aims at enabling wide cost-efficient implementation of sensing solutions for surface recognition via walking-pattern classification by restricting the necessary hardware to a cost-economic microprocessor and a single type of force sensors. For experimental analysis, we explored the walking-pattern classification performance using a framework which combines a support vector machine (SVM) and four time-domain feature descriptors, i.e., mean of amplitude (MA), integral of absolute value (IAV), variance (VAR), and root mean square (RMS). During the online pattern classification, the dynamical force-sensory-data stream was extracted using a real-time overlapped-window-based method. Multiple binary SVM classifiers were applied for solving the multi-class classification problem, due to the reasonably high accuracy and the relatively small complexity for hardware implementation, allowing simultaneous strength exploitation of above four individual feature descriptors with a one-versus-one (OVO) strategy. The experimental results, obtained with 250 samples/surface, verified 93.8% mean average precision, 93.7% average accuracy and recall rates of 98.8%, 91.6%, 82.0%, 98.0%, 98.0% for smooth wood, rough foam, smooth foam, thick carpet, and thin carpet, respectively. Only the dynamical force-sensing data were employed for a 10-fold cross validation, which enabled the high processing speed of 0.73 ms/stride. The developed cost-efficient and accurate surface-recognition system can be useful for ensuring safe in-door locomotion for the biped robot and can help the robot to better understand the human living environment by increasing its sensing diversity.

Machine learning aided carrier recovery in continuous-variable quantum key distribution

The secret key rate of a continuous-variable quantum key distribution (CV-QKD) system is limited by excess noise. A key issue typical to all modern CV-QKD systems implemented with a reference or pilot signal and an independent local oscillator is controlling the excess noise generated from the frequency and phase noise accrued by the transmitter and receiver. Therefore accurate phase estimation and compensation, so-called carrier recovery, is a critical subsystem of CV-QKD. Here, we explore the implementation of a machine learning framework based on Bayesian inference, namely an unscented Kalman filter (UKF), for estimation of phase noise and compare it to a standard reference method and a previously demonstrated machine learning method. Experimental results obtained over a 20-km fibre-optic link indicate that the UKF can ensure very low excess noise even at low pilot powers. The measurements exhibited low variance and high stability in excess noise over a wide range of pilot signal to noise ratios. This may enable CV-QKD systems with low hardware implementation complexity which can seamlessly work on diverse transmission lines.

A Controlled Accuracy-Based Recursive Algorithm for Approximation of Sigmoid Activation

The sigmoid activation function is widely used in the development of neural networks. It is implemented in hardware development using an approximation scheme. In this paper, a piecewise linear (PWL) approximation scheme is developed as an accuracy controlled recursive algorithm. It uses a maximum allowable error as an input parameter and the least square method for finding the PWL function. The algorithm’s accuracy is calculated with an absolute average and maximum error and compared to the state of the art. The approximation scheme’s hardware implementation complexity is compared in terms of the number of PWL functions required. Results show that the proposed algorithm requires comparatively fewer PWL functions for the approximate sigmoid function for the specified accuracy.

Addressing variability in reuse prediction for last-level caches

Last-Level Cache (LLC) represents the bulk of a modern CPU processor's transistor budget and is essential for application performance as LLC enables fast access to data in contrast to much slower main memory. However, applications with large working set size often exhibit streaming and/or thrashing access patterns at LLC. As a result, a large fraction of the LLC capacity is occupied by dead blocks that will not be referenced again, leading to inefficient utilization of the LLC capacity. To improve cache efficiency, the state-of-the-art cache management techniques employ prediction mechanisms that learn from the past access patterns with an aim to accurately identify as many dead blocks as possible. Once identified, dead blocks are evicted from LLC to make space for potentially high reuse cache blocks. In this thesis, we identify variability in the reuse behavior of cache blocks as the key limiting factor in maximizing cache efficiency for state-of-the-art predictive techniques. Variability in reuse prediction is inevitable due to numerous factors that are outside the control of LLC. The sources of variability include control-flow variation, speculative execution and contention from cores sharing the cache, among others. Variability in reuse prediction challenges existing techniques in reliably identifying the end of a block's useful lifetime, thus causing lower prediction accuracy, coverage, or both. To address this challenge, this thesis aims to design robust cache management mechanisms and policies for LLC in the face of variability in reuse prediction to minimize cache misses, while keeping the cost and complexity of the hardware implementation low. To that end, we propose two cache management techniques, one domain-agnostic and one domain-specialized, to improve cache efficiency by addressing variability in reuse prediction.

Detection of epileptical seizures based on alpha band statistical features

Significant research has been going in the field of automated epileptical seizure detection using Electroencephalogram (EEG) data. The EEG signal consists of different frequency bands, which correspond to the different emotional and mental activities of the humans. Most of the research work uses the whole frequency spectrum for the detection of seizures. In this paper, first time the proposed automated system utilizing machine learning technique using only alpha band (8–12 Hz). This paper uses Short-time Fourier transform (STFT) due to its high speed and less complexity in hardware implementation to convert EEG data in time–frequency (t–f) plane. As brain oscillations of a person vary in different health conditions, four statistical features have been extracted from t–f plane of alpha band. The detection performance of the features of alpha band has been analyzed on six classifiers using tenfold cross-validation which shows that the Random Forest (RF) classifier gives the best performance among different classifiers for most of the experiments performed. This study has achieved the best classification accuracy of 98% and ROC analysis revealed maximum Area Under Curve (AUC) of 1 to distinguish the seizures and healthy. Hence, the statistical features of the alpha band depict to be a potential biomarker for the real time detection system.

P-SFA: Probability based Sigmoid Function Approximation for Low-complexity Hardware Implementation

A probability-based sigmoid function approximation (P-SFA), which is based on piecewise linear function and neuron's values statistical probability distribution in each layer, is proposed to lower the complexity of neural network hardware implementation with only addition circuit. The sigmoid function is divided into three fixed regions, and the number of sub-regions with different sizes in each fixed region is adapted to neuron's values distribution in each layer to reduce the error between the sigmoid function and P-SFA function. The experimental results on FPGA show that the P-SFA function is efficient in terms of power and speed, and the recognition accuracies in DNN and CNN for MNIST with P-SFA are the highest among the state-of-the-art methods, up to 97.46% and 99.02% respectively.

A model of parallel sorting neural network of discrete-time

A model of parallel sorting neural network of discrete-time is presented. The model is described by a system of differential equations and by step functions. The network has high speed, any finite resolution of input data and it can process unknown input data of finite values located in arbitrary finite range. The network is characterized by moderate computational complexity and complexity of hardware implementation. The results of computer simulation illustrating the efficiency of the network are provided.

Hardware Implementation of Parallelized Fuzzy Adaptive Resonance Theory Neural Network

A hardware implementation design of parallelized fuzzy Adaptive Resonance Theory neural network is described and simulated. Parallel category choice and resonance are implemented in the network. Continuous-time and discrete-time winner-take-all neural circuits identifying the largest of M inputs are used as the winner-take-all units. The continuous-time circuit is described by a state equation with a discontinuous right-hand side. The discrete-time counterpart is governed by a difference equation. Corresponding functional block-diagrams of the circuits include M feed-forward hard- limiting neurons and one feedback neuron, which is used to compute the dynamic shift of inputs. The circuits combine arbitrary finite resolution of inputs, high convergence speed to the winner-take-all operation, low computational and hardware implementation complexity, and independence of initial conditions. The circuits are also used for finding elements of input vector with minimal/maximal values to normalize them in the range [0,1].

Detecting Object Open Angle and Direction Using Machine Learning

Nowadays, the object detection techniques have been developed rapidly for different applications, ranging from remote sensing to autonomous vehicles. We demonstrate identification of object open angle and direction using machine learning (ML) algorithms based on received light beam’s intensity profiles. Compared with previous optical orbital-angular-momentum (OAM) spectrum system and other related works, our proposed technique only uses a single-shot image, and can efficiently reduce the complexity of hardware implementation. Specifically, we verify the reliability of the simulation results experimentally for 14 open angles and 32 directions. Experimental result shows that convolutional neural network (CNN) outperforms the other traditional ML algorithms, such as decision tree (DT), k-nearest neighbor algorithm (KNN), and support vector machine (SVM). As one of the variant of CNN, MobileNet (MN) has relatively simplified iteration algorithm than VGG-like net. It reduces the computational power, while still maintaining high accuracy for identification issues.

Application of compressive sensing based on multimode fiber specklegram in optical image encryption

In order to ensure the secure and effective transmission of image information, a new method of optical image encryption using the multimode fiber (MMF) specklegram based compressive sensing combined with the double random phase encoding (DRPE) is proposed in this paper. The specklegrams obtained from the facet of the multimode fiber are used as the measurement matrix of compressive sensing (CS), and the compression and the first-stage encryption of the image are completed by compressive sensing, in which the specklegram also functions as the first secret key. Then, the second-stage encryption is implemented by using the double random phase encoding technology, in which the random phase mask acts as the second secret key. All of the specklegrams used in this paper are obtained from the facet of a 5 m-long and 105-μm-diameter-MMF and offset launching technique. Then the fiber specklegrams are proposed in several steps to provide the measurement matrix in CS. By performing an encryption and decryption test on a standard Lena image of 256 × 256 size, it is found that the decrypted image and the original image are visually consistent, and the encryption is also realized in the process of compression, which indicates the method proposed in this paper is feasible. Furthermore, the comparison studies of the performances of specklegram based measurement matrix and some classic measurement matrices show that the decrypted image quality using the specklegram matrix is better. And at the same time, comparing with the high hardware implementation complexity and high cost of other measurement matrices, specklegram based matrix can be easily realized by simple optical device, and the corresponding secret key can be easily changed by the working wavelength, which is helpful for enlarging the secret key space. It is further proved that the encryption method be able to effectively resist the statistical analysis attacks, cropping attacks and noise interference, and also have high sensitivity to the secret key, which shows good robustness and high security. Therefore, the image encryption method combined with the specklegram matrix based compression sensing with the optical DRPE can obtain good encryption effect and has a great secret key space, which may provide a good candidate scheme for the pure optical realization of image encryption.