Look-up Table Research Articles

An optimal area power system-based automatic load frequency control using a 1-D lookup table method

An optimal area power system-based automatic load frequency control using a 1-D lookup table method

BDLUT: Blind image denoising with hardware‐optimized look‐up tables

AbstractDenoising sensor‐captured images on edge display devices remains challenging due to deep neural networks' (DNNs) high computational overhead and synthetic noise training limitations. This work proposes BDLUT(‐D), a novel blind denoising method combining optimized lookup tables (LUTs) with hardware‐centric design. While BDLUT describes the LUT‐based network architecture, BDLUT‐D represents BDLUT trained with a specialized noise degradation model. Designed for edge deployment, BDLUT(‐D) eliminates neural processing units (NPUs) and functions as a standalone ASIC IP solution. Experimental results demonstrate BDLUT‐D achieves up to 2.42 dB improvement over state‐of‐the‐art LUT methods on mixed‐noise‐intensity benchmarks, requiring only 66 KB storage. FPGA implementation shows over 10 reduction in logic resources, 75% less storage compared to DNN accelerators, while achieving 57% faster processing than traditional bilateral filtering methods. These optimizations enable practical integration into edge scenarios like low‐cost webcam enhancement and real‐time 4 K‐to‐4 K denoising without compromising resolution or latency. By enhancing silicon efficiency and removing external accelerator dependencies, BDLUT(‐D) establishes a new standard for practical edge imaging denoising. Implementation is available at https://github.com/HKU-LiBoyu/BDLUT.

Transforming Group Codes in Mealy Finite State Machines with Composite State Codes

A new state assignment method focusing on Mealy finite state machines (FSMs) is proposed. The proposed codes are an alternative to composite state codes (CSCs). CSCs are represented as concatenations of group codes and partial state codes. Both group and partial state codes are maximum binary codes. We propose encoding groups using one-hot codes. The main goal of this method is improving the value of the FSM cycle time without a significant degradation of the spatial characteristics. The method can be applied if FSM circuits are implemented using the look-up table (LUT) elements of field-programmable gate arrays (FPGAs). The resulting FSM circuit includes three logic blocks. The first block generates partial input memory functions and FSM outputs depending on maximum binary state codes and one-hot group codes. The partial codes are assigned in a way minimizing the number of arguments in the partial functions. This allows for the generation of most partial functions by single-LUT circuits. The second block generates the final values of the input memory functions and FSM outputs. This block does not require group codes to generate functions, as in CSC-based FSMs. The third block transforms maximum binary group codes into their one-hot equivalents. The proposed approach allows for a reduction in the number of series-connected LUTs in comparison with CSC-based FSMs. Due to this reduction, the temporal characteristics of an FSM circuit are improved. This paper includes an example of FSM synthesis applying the proposed method. The experiments were conducted using standard benchmark FSMs. The results of the experiments show that the proposed method allowed for an improvement in the cycle time of an average of 8.81%. Moreover, in relation to CSC-based FSMs, the LUT counts decreased by an average of 4.00%.

Research on Effective Radius Retrievals of Aerosol Particles Based on Dual-Wavelength Lidar

In this study, the effective radius of aerosol particles was experimentally retrieved using a self-developed dual-wavelength atmospheric aerosol lidar. A single-valued lookup table was first established, based on the OPAC database and the Gamma size distribution model, to define the relationship between the extinction coefficient ratio and the effective radius of atmospheric aerosol particles. The extinction coefficients corresponding to the 355 nm and 1064 nm wavelengths were then calculated using the echo signals retrieved horizontally by the lidar, in conjunction with the Mie scattering lidar equation. Subsequently, the lookup table was used to retrieve the real-time effective radius of aerosol particles by inputting the extinction coefficient ratio of the two wavelengths. Finally, the retrieval results were compared with the effective radii measured by an optical particle spectrometer, which had been corrected for relative humidity. An analysis over six months showed a coefficient of determination (R2) greater than 0.83. The results demonstrated that the dual-wavelength lidar exhibits a stable performance, the retrieval method is valid, and the detection results are accurate and reliable.

SHLUT: Efficient Image Enhancement using Spatial‐Aware High‐Light Compensation Look‐up Tables

AbstractRecently, the look‐up table (LUT)‐based method has achieved remarkable success in image enhancement tasks with its high efficiency and lightweight nature. However, when considering edge scenarios with limited computational resources, most existing methods fail to meet practical requirements due to their costly floating‐point operations on convolution layers, which limit their general use. Moreover, most LUT‐based methods may not perform well in handling high‐light regions. To address these issues, we propose SHLUT, an efficient and practical image enhancement method by using spatial‐aware high‐light compensation look‐up tables (LUTs), which comprise two parts. Firstly, we propose a spatial‐aware weight predictor to reduce the computational burden. A lightweight network is trained to predict spatial‐aware weight values, and then we transfer the values to the LUTs. Additionally, to correct overexposure in high‐light regions, we propose a high‐light compensation 3D LUT. Our proposed method allows us to directly retrieve the values from the LUTs to achieve efficient image enhancement at test time. Extensive experimental results demonstrate that SHLUT exhibits competitive performance compared to other LUT‐based methods both quantitatively and qualitatively in a more efficient manner. For instance, SHLUT significantly reduces computational resources (at least 18 times in GFLOPs compared to other LUT‐based methods), while excelling in high‐light region handling.

Wireless Charger for Pacemakers Controlled from Primary Current Without Communication with Secondary Side

This paper discusses the implementation of a wireless inductive power transfer system for pacemaker applications. One of the inherent challenges in these systems is regulating the output voltage, as there is no direct physical connection from the primary. Additionally, there are other challenges, such as variability in magnetic coupling. First, resonant converters for inductive charging topologies are investigated for biomedical applications. Then, a control method based on the system’s modeling is proposed, eliminating the need for communication. This method is designed for systems with variable and unknown coupling and specifically for a resonant series–parallel topology. For an operation point, determined by the coupling factor, the primary current is measured to regulate the output voltage by adjusting the input voltage. The relationship between the input current and the input voltage is set by a look-up table. The effectiveness of this control strategy is validated in the PSIM simulator and with experimental results for a coupling range between 0.3 and 0.5, achieving a regulated output current error of less than 1%, and an output voltage range within the limits of the battery charger.

Creating a common scale for the conversion of qDASH and PRWHE scores.

Creating a common scale for the conversion of qDASH and PRWHE scores.

Efficient Neural Network Encoding for 3D Color Lookup Tables

3D color lookup tables (LUTs) enable precise color manipulation by mapping input RGB values to specific output RGB values. 3D LUTs are instrumental in various applications, including video editing, in-camera processing, photographic filters, computer graphics, and color processing for displays. While an individual LUT does not incur a high memory overhead, software and devices may need to store dozens to hundreds of LUTs that can take over 100 MB. This work aims to develop a neural network architecture that can encode hundreds of LUTs in a single compact representation. To this end, we propose a model with a memory footprint of less than 0.25 MB that can reconstruct 512 LUTs with only minor color distortion (ΔE ≤ 2.0 on average) over the entire color gamut. We also show that our network can weight colors to provide further quality gains on natural image colors (ΔE ≤ 1.0 on average). Finally, we show that minor modifications to the network architecture enable a bijective encoding that produces LUTs that are invertible, allowing for reverse color processing.

DPLUT: Unsupervised Low-light Image Enhancement with Lookup Tables and Diffusion Priors

Low-light image enhancement (LIE) aims at precisely and efficiently recovering an image degraded in poor illumination environments. Recent advanced LIE techniques are using deep neural networks, which require lots of low-normal light image pairs, network parameters, and computational resources. As a result, their practicality is limited. In this work, we devise a novel unsupervised LIE framework based on diffusion priors and lookup tables (DPLUT) to achieve efficient low-light image recovery. The proposed approach comprises two critical components: a light adjustment lookup table (LLUT) and a noise suppression lookup table (NLUT). LLUT is optimized with a set of unsupervised losses. It aims at predicting pixel-wise curve parameters for the dynamic range adjustment of a specific image. NLUT is designed to remove the amplified noise after the light brightens. As diffusion models are sensitive to noise, diffusion priors are introduced to achieve high-performance noise suppression. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods in terms of visual quality and efficiency.

Pushing the Limits of BFP on Narrow Precision LLM Inference

The substantial computational and memory demands of Large Language Models (LLMs) hinder their deployment. Block Floating Point (BFP) has proven effective in accelerating linear operations, a cornerstone of LLM workloads. However, as sequence lengths grow, nonlinear operations, such as Attention, increasingly become performance bottlenecks due to their quadratic computational complexity. These nonlinear operations are predominantly executed using inefficient floating-point formats, which renders the system challenging to optimize software efficiency and hardware overhead. In this paper, we delve into the limitations and potential of applying BFP to nonlinear operations. Given our findings, we introduce a hardware-software co-design framework (DB-Attn), including: (i) DBFP, an advanced BFP version, overcomes nonlinear operation challenges with a pivot-focus strategy for diverse data and an adaptive grouping strategy for flexible exponent sharing. (ii) DH-LUT, a novel lookup table algorithm dedicated to accelerating nonlinear operations with DBFP format. (iii) An RTL-level DBFP-based engine is implemented to support DB-Attn, applicable to FPGA and ASIC. Results show that DB-Attn provides significant performance improvements with negligible accuracy loss, achieving 74% GPU speedup on Softmax of LLaMA and 10x low-overhead performance improvement over SOTA designs.

Development of a smartphone-based multi-spectral imaging system and its performance validation through radiodermatitis assessment

Cancerous skin lesions (i.e., skin cancers) are the most common types of cancer. Currently, there exists an unmet clinical need to develop advanced imaging tools for skin diagnosis and associated treatment monitoring. In this work, we reported the development of a smartphone-based multi-spectral imaging (MSI) system to fulfill the need and validated its performance through radiodermatitis assessment: one of the severe side effects of skin cancer radiotherapy (RT) treatment that needs to be objectively graded to avoid the possible RT treatment cessation thereof. In combination with a graphics processing unit (GPU)-accelerated Monte Carlo simulations code, Monte Carlo look-up table (MCLUT)-based skin-layer-resolved physiological parameters [i.e., the blood volume fraction (BVF) and the oxygen saturation of hemoglobin in blood] were derived from the multi-spectra measured. The mean square error (MSE) between the MCLUT-derived spectra and the experimentally acquired multi-spectra was found to be less than 0.0031, confirming the robustness of the physiological parameters derived. Further analysis of variance (ANOVA) conducted on the BVF and the oxygen saturation of hemoglobin in blood revealed a significant (P<0.001, ANOVA) increase in erythema. Additionally, partial least square discriminant analysis (PLS-DA) implemented on the multi-spectra leads to an overall classification accuracy of 85.9% for the separation between normal skin and erythema, confirming the potential of the smartphone-based MSI system developed for radiodermatitis assessment.

Simplified control of neuromuscular stimulation systems for restoration of reach with limb stiffness as a modifiable degree of freedom.

Brain-controlled functional electrical stimulation (FES) of the upper limb has been used to restore arm function to paralyzed individuals in the lab. Able-bodied individuals naturally modulate limb stiffness throughout movements and in anticipation of perturbations. Our goal is to develop, via simulation, a framework for incorporating stiffness modulation into the currently-used 'lookup-table-based' FES control systems while addressing several practical issues: 1) optimizing stimulation across muscles with overlap in function, 2) coordinating stimulation across joints, and 3) minimizing errors due to fatigue. Our calibration process also needs to account for when current spread causes additional muscles to become activated.
Approach: We developed an analytical framework for building a lookup-table-based FES controller and simulated the clinical process of calibrating and using the arm. A computational biomechanical model of a human paralyzed arm responding to stimulation was used for simulations with six muscles controlling the shoulder and elbow in the horizontal plane. Both joints had multiple muscles with overlapping functional effects, as well as biarticular muscles to reflect complex interactions between joints. Performance metrics were collected in silico, and real-time use was demonstrated with a Rhesus macaque using its cortical signals to control the computational arm model in real time.
Main Results: By explicitly including stiffness as a definable degree of freedom in the lookup table, our analytical approach was able to achieve all our performance criteria. While using more empirical data during controller parameterization produced more accurate lookup tables, interpolation between sparsely sampled points (e.g., 20 degree angular intervals) still produced good results with median endpoint position errors of less than 1 cm-a range that should be easy to correct for with real-time visual feedback.
Significance: Our simplified process for generating an effective FES controller now makes translating upper limb FES systems into mainstream clinical practice closer to reality. &#xD.

Low-Error ASIC Implementation of SoftMax Activation Function for Deep Neural Networks

Deep Neural Networks (DNNs) are one of the prominent and state-of-the-art methods for implementing artificial intelligence algorithms. A noticeable amount of effort has been put into hardware acceleration of DNN, with significantly less attention given to the SoftMax layer, which uses expensive hardware for exponentiation and division. The SoftMax function must be implemented as accurately as possible since it is crucial in multiclass classification training and inference tasks. This paper describes an efficient implementation of the SoftMax function for error-sensitive application of DNNs. We propose a Lookup Table (LUT) based on linear polynomial approximation for exponential and log units. The proposed architecture is modeled in Verilog and synthesized to gate-level netlist in GPDK180[Formula: see text]nm using Cadence[Formula: see text]. Results after the placement and routing stage show a delay of 7.15[Formula: see text]ns while consuming 5.3 mW of total power when operated at 100[Formula: see text]MHz. Multiclass classification is coded in Python with the proposed SoftMax layer to evaluate the accuracy in real-time applications. The results show that the neural network trained with the proposed SoftMax reaches almost the same accuracy as traditional SoftMax of 86.76%. This proves that low-form factor platforms can implement the same accuracy achieved with the proposed SoftMax function with less power.

Implementation of lookup tables for different optimization strategies of semi-active car suspension system

Road irregularities and various vehicle loads influence comfort and safety levels. Owing to these changes, the driver cannot quickly and easily find the best driving parameters. Control of damping in a semi-active suspension adjusts the damping process in the vehicle to minimize the acceleration of the crew. This ensures comfort for them, influencing the level of fatigue of the driver and safe driving. A theoretical analysis was implemented using a mathematical full-car model in Simulink/MATLAB. We performed a simulation of a vehicle with all passengers passing various artificially generated road profiles at different velocities. We optimized the damping coefficient for the maximum comfort level using one, two, or four damping values, implementing different optimization strategies. The obtained research results were finalized by the conclusions.

Arrhythmia classification using CMF-AFF based on electrocardiogram in field programmable gate array device

Arrhythmia classification is categorization of irregular heart rhythms depending on patterns detected in electrocardiogram (ECG) signals assist in treatment and diagnosis of cardiac conditions. ECG evaluates heart’s electrical activity to diagnose various heart conditions, but it is affected by interference or noise. ECG’s signal filtering is essential pre-processing stage that minimizes noise and highlights wave characteristics in ECG data. However, digital filters are normally constructed by multiplying coefficient and then multiplying value given as feedback which leads to more power and area consumption. To solve these issues, coefficient memory compression (CMC) technique is proposed with an adaptive FIR filter (AFF) to achieve low area and low power dissipation by compressing memory requirements in a field programmable gate array (FPGA). An adaptive FIR filter is employed to effectively minimize noise like baseline noise, muscle contraction noise, and low-frequency noise. The performance of CMC-AFF is analyzed in terms of look up table (LUT), register, digital signal processing (DSP), power, and global buffer (BufG). The proposed approach achieves a low power consumption of 0.012 W in Zed Board Zynq7000 AP system on chip (SoC) FPGA device compared to existing techniques like collateral and sequence approaches using Bartlet filter and low-power ECG processor using Bartlet filter respectively.

Analysis of VFDPC for three-level neutral point clamped AC-DC converters with capacitor balancing s

This paper presents an analysis of the dynamic performance of a three-level neutral point clamped (NPC) AC-DC converter utilizing the advanced control technique of virtual flux direct power control (VFDPC). VFDPC estimates the three-phase grid voltage and instantaneous active and reactive power components, eliminating the need for an AC input voltage sensor used in conventional direct power control (DPC). This reduction in sensors decreases system complexity and cost while mitigating high-frequency noise and interference. Integrating VFDPC into 3L NPC AC-DC converters significantly enhances overall performance, leading to more efficient and robust power conversion systems. However, a significant challenge in the three-level NPC topology is the voltage imbalance in the neutral point of the DC-link capacitor, which can cause excessive voltage stress on switching devices and degrade system performance. To address this, a novel lookup table has been developed, incorporating strategies to balance the capacitor voltage. The results of this study demonstrate that VFDPC generates nearly sinusoidal line currents with reduced current total harmonic distortion (THD). Additionally, VFDPC ensures unity, lagging, and leading power factor operation, while providing flexibility to adjust the DC-link output voltage and accommodate load variations. These capabilities highlight VFDPC effectiveness in managing power quality and system stability, even under varying load conditions.

Technical note: Extension of full-spectrum correlated k-distribution look-up tables to CH4 and NH3 for use in combustion modelling

Technical note: Extension of full-spectrum correlated k-distribution look-up tables to CH4 and NH3 for use in combustion modelling

Improving Temporal Characteristics of Mealy FSM with Composite State Codes

In this paper, we proposed a new state assignment method focusing on Mealy finite state machines (FSMs). The method makes it possible to improve the temporal characteristics of the circuits of FSMs, the internal states of which are encoded by the composite state codes (CSCs). These codes consist of class codes and partial state codes. Both class and partial state codes are maximum binary codes. We propose to encode classes by one-hot codes. The main goal of the method is improving the value of the FSM cycle time without any significant degradation of spatial characteristics. The method can be applied if FSM circuits are implemented using look-up table (LUT) elements of field-programmable gate arrays (FPGAs). The resulting FSM circuit includes two logic blocks. The first block generates partial input memory functions and FSM outputs depending on maximum binary state codes and one-hot class codes. The choice of partial codes allows minimizing the systems of partial functions. This allows generating most partial functions by single-LUT circuits. Some partial functions require using dedicated multiplexers. The second block generates final values of input memory functions and FSM outputs. This block does not require class codes to generate functions, which is the case of CSC-based FSMs. The proposed approach allows reducing the number of series-connected LUTs in comparison with CSC-based FSMs. Due to this reduction, the temporal characteristics are improved. The paper includes an example of FSM synthesis through applying the proposed method. The experiments are conducted using standard benchmark FSMs. The results of experiments show that the proposed method allows improving the temporal characteristics (by an average of 9.15%). In relation to CSC-based FSMs, the number of LUTs increases by an average of 10.03%, and the power consumption increases by an average of 7.63%.

A quantum random access memory (QRAM) using a polynomial encoding of binary strings

Quantum algorithms claim significant speedup over their classical counterparts for solving many problems. An important aspect of many of these algorithms is the existence of a quantum oracle, which needs to be implemented efficiently in order to realize the claimed advantages in practice. A quantum random access memory (QRAM) is a promising architecture for realizing these oracles. In this paper we develop a new design for QRAM and implement it with Clifford+T circuit. We focus on optimizing the T-count and T-depth since non-Clifford gates are the most expensive to implement fault-tolerantly in most error correction schemes. Integral to our design is a polynomial encoding of bit strings and so we refer to this design as . Compared to the previous state-of-the-art bucket brigade architecture for QRAM, we achieve an exponential improvement in T-depth, while reducing T-count and keeping the qubit-count same. Specifically, if N is the number of memory locations to be queried, then has T-depth , T-count and uses O(N) logical qubits, while the bucket brigade circuit has T-depth , T-count O(N) and uses O(N) qubits. Combining two we design a quantum look-up-table, , that has T-depth , T-count and qubit count . A quantum look-up table (qLUT) or quantum read-only memory (QROM) has restricted functionality than a QRAM. For example, it cannot write into a memory location and the circuit needs to be compiled each time the contents of the memory change. The previous state-of-the-art CSWAP architecture has T-depth , T-count and qubit count . Thus we achieve a double exponential improvement in T-depth while keeping the T-count and qubit-count asymptotically same. Additionally, with our polynomial encoding of bit strings, we develop a method to optimize the Toffoli-count of circuits, specially those consisting of multi-controlled-NOT gates.

Computerized Mind-Reading Magic Trick with Remote Transmission

Magic is typically presented live, and computer presentation is relatively rare. This article is based on mind-reading magic and presents the magic process on a computer. Twenty-one cards are given to the spectator to make choices. Through the processes of dealing, collecting and questioning, the playing cards considered are transformed through mathematical derivation. This article uses the tree table lookup method to replace the mathematical algorithm used in the original magic. The program execution results can indeed achieve good results. Finally, we considered only displaying the card that the person was thinking about on the screen, but found this too monotonous and lacking in interactivity. Thus, the magic result will be presented on a touch panel or mobile webpage, so as to set off the final climax at the end of the magic. The playing card suits and numbers generated by the computer are transmitted to the touch panel via Bluetooth. Another method is to use Raspberry Pi to set up a server and create a database form, sending the card suits and numbers to the database form through Wi-Fi. Players can then connect to the IP address through their mobile phones to display the picture of the playing cards.