Low Computational Complexity Research Articles

Reducing Peak Power in a Multiple Load System by Delaying the Activation of Electrical Loads Using a Filter Based on a PI Controller

In a power grid with multiple two-state loads, the total power can vary over a significant range. This results in the inability to supply this system from a low-power source. The solution is an algorithm that shapes energy demand depending on its availability. For this purpose, a new load distribution method is proposed based on introducing a buffer between the temperature controller output and the heater and filtering the load using a master Proportional–Integral (PI) controller. The aim of the work was to evaluate the quality of the developed algorithm for limiting power peaks in the power grid. The research was conducted on a model of the Creep Test Laboratory with 389 heaters simulated in MATLAB Simulink R2023b. The algorithm was tested with various settings of the master controller parameters. By experimentally adjusting these parameters, a ten-fold reduction in peak power was achieved. The standard deviation for the L1 phase was reduced from 7.6 kW to 0.6 kW. Similar results were obtained for phases L2 and L3. The tested control system tracked changes in the average power value by changing the number of loads switched on and by frequency-modulating the signal when the change was less than the power of a single load. It was demonstrated that the controlled delayed switching of electrical loads can modify the shape of the total electric power without affecting their operation. The proposed solution features a low computational complexity, which allows its implementation in various systems.

Emissions and Flame Stability Assessment of Hydrogen Addition and Air Dilution in a Micro Gas Turbine Combustor Using 0D/1D Modeling by Chemical Reactor Network

Abstract Hydrogen emerges as a promising fuel for clean and sustain- able electricity production when utilized in micro gas turbines (mGTs). However, some challenges linked to hydrogen usage must be overcome as its high reactivity can lead to increased NOx emissions and flame instabilities. Literature suggests that combustion air humidification and/or exhaust gas recirculation (EGR) may be methods to reduce this reactivity. However, these measures are limited due to the small operating range of the mGT combustor. In this context, a 20 kWth mGT combustor is investigated un- der various inlet conditions to assess the effect of EGR towards increased flame stability and emission control when operating under hydrogen-enriched methane firing, using a Chemical Re- action Network (CRN) model. The CRN modeling is a fast and low computational complexity tool to model combustion perfor- mance and emissions by representing complex reactive flow fields through idealized reactor models, leading to reduced computa- tional costs. This study examines partial load conditions, fuel compositions, and the effect of combustion air dilution through EGR. The CRN model of the combustion process is designed based on combustion zones extracted from the main flow fields with similar thermo-chemical states from CFD, while emission predictions are experimentally validated for different operating points. Predicted NOx and CO emissions by the proposed CRN model show an acceptable agreement with the experimental data at high power loads. However, its accuracy diminishes for loads below 70% due to the variability in model tuning parameters across different loads, whereas the

Intelligent cell images segmentation system: based on SDN and moving transformer

Diagnosing diseases heavily relies on cell pathology images, but the extensive data in each manual identification of relevant cells labor-intensive, especially in regions with a scarcity of qualified healthcare professionals. This study aims to develop an intelligent system to enhance the diagnostic accuracy of cytopathology images by addressing image noise and segmentation issues, thereby improving the efficiency of medical professionals in disease diagnosis. We introduced an innovative system combining a self-supervised algorithm, SDN, for image denoising with data enhancement and image segmentation using the UPerMVit model. The UPerMVit model’s novel attention mechanisms and modular architecture provide higher accuracy and lower computational complexity than traditional methods. The proposed system effectively reduces image noise and accurately segments annotated images, highlighting cellular structures relevant to medical staff. This enhances diagnostic accuracy and aids in the accurate identification of pathological cells. Our intelligent system offers a reliable tool for medical professionals, improving diagnostic efficiency and accuracy in cytopathologic image analysis. It provides significant technical support in regions lacking adequate medical expertise.

Research on position tracking control of electro-hydraulic actuator system combining feedforward inverse model controller and exponential functional link network adaptive controller

The electro-hydraulic actuator system plays a crucial role in indoor high-frequency simulation experiments. Many components of the system exhibit unavoidable nonlinear characteristics, which reduce its dynamic position response capability. Additionally, most algorithms are unsuitable for practical high-frequency conditions due to either local instability or high computational complexity. To improve position tracking accuracy under high-frequency control, a novel combined control strategy is proposed, integrating an appropriate feedforward inverse model controller (FFIMC) with an exponential functional link network adaptive controller (EFLNAC). Initially, the system has reached a stable state under the action of the PID controller. Subsequently, the FFIMC estimates both the model and the inverse model of the PID system by applying the AIC technique developed through the DCT-VSSNLMS adaptive algorithm, which extends the system’s bandwidth. Following this, the EFLNAC, as the outermost loop, acquires the inverse model of the FFIMC system and updates internal parameters in real-time, thereby enhancing the system’s resistance to nonlinear disturbances. Finally, the combined controller is validated through simulations and experiments under various working conditions. The results show a significant increase in the PID system’s bandwidth and demonstrate that the tracking error of the signal is maintained within 0.079 mm. Therefore, the combined controller proves to be an effective solution for high-precision, high-frequency displacement tracking, offering the advantages of low computational complexity, high accuracy, and strong effectiveness.

Optimized Accelerated Over-Relaxation Method for Robust Signal Detection: A Metaheuristic Approach

Massive MIMO technology is recognized as a key enabler for beyond 5G (B5G) and next-generation wireless networks. By utilizing large-scale antenna arrays at the base station (BS), it significantly improves both system capacity and energy efficiency. Despite these advantages, the deployment of a high number of antennas at the BS presents considerable challenges, particularly in the design of signal detectors that can operate with low computational complexity. While the minimum mean square error (MMSE) detector offers optimal performance in these large-scale systems, it suffers from the computational burden that makes its practical implementation challenging. To mitigate this, various iterative methods and their improved versions have been introduced. However, these iterative methods often converge slowly and are less accurate. To address these challenges, this study introduces an improved variant of traditional accelerated over-relaxation (AOR), called optimized AOR (OAOR). AOR is an over-relaxation method, and its performance is highly dependent on its relaxation parameters. To find the optimal parameters, we have developed an innovative approach that integrates a nature-inspired meta-heuristic algorithm known as Particle Swarm Optimization (PSO). Specifically, we introduce a novel variant of PSO that improves upon basic PSO by enhancing the cognitive coefficients to optimize the relaxation parameters for OAOR. These key modifications to the standard PSO improve its ability to explore various solutions efficiently and help to find the optimal parameters more quickly for signal detection. It facilitates the OAOR with faster convergence towards the optimal solution by reducing the error rate, resulting in high detection accuracy and simultaneously decreasing computational complexity from O(K3) to O(K2) making it suitable for modern wireless communication systems. We conduct extensive simulations across various configurations of massive MIMO systems. The results indicate that our proposed method achieves better performance compared to existing techniques. This improvement is particularly evident in terms of both computational complexity and error rate.

Optimizing size and location of UPFC for enhanced system dynamic stability using hybrid approach

This paper proposed a novel hybrid optimization technique, combining the Enhanced Multi-Strategic Sparrow Search Algorithm (EMSA) and White Shark Optimizer (WSO), to determine the optimal location and size of a Unified Power Flow Controller (UPFC) for enhancing power system dynamic stability. The proposed method offers improved search capabilities, reduced randomness, and lower computational complexity compared to existing approaches. Generator faults can significantly impact system dynamic stability constraints, including voltage and power loss. EMSA algorithm is employed to identify optimal location for UPFC placement by selecting bus with minimum power loss. Subsequently, WSO algorithm is used to optimize UPFC's capacity, ensuring that affected system parameters and dynamic stability constraints are restored within safe limits. Optimized UPFC is then installed at identified location, and system's power flow is analyzed. Proposed method is implemented in MATLAB/Simulink environment and tested on both IEEE 30 and IEEE 14 standard benchmark systems. Proposed method's performance is evaluated by comparison with existing methods.

Diffusion Augmented Complex Inverse Square Root for Adaptive Frequency Estimation over Distributed Networks

Using adaptive filtering to estimate the frequency of power systems has become a popular trend. In recent years, however, few studies have been performed on adaptive frequency estimations in non-stationary noise environments. In this paper, we propose the distributed complex inverse square root algorithm and distributed augmented complex inverse square root algorithm for the frequency estimation of power systems based on the widely linear model and the inverse square root cost function, where the function can restrain both positive and negative large errors, based on its symmetry. Moreover, the wireless sensor networks support monitoring and adaptation for the frequency estimation in the distributed networks, and the proposed approach can ensure good robustness of the balanced or unbalanced three-phase power system with the help of a local complex-value voltage signal generated by Clark’s transformation. In addition, the bound of step size is driven by the global vectors, and that low computation complexity do not hinder those performances. The results of several experiments demonstrate that our algorithms can effectively estimate the frequency in impulsive noise environments.

Short-time adaptive compact kernel distribution for fault diagnosis under variable working condition bearing

Time-frequency distribution (TFD) methods are extensively employed in the diagnosis of bearings operating under variable working conditions. However, these methods often face high computational complexity, limited time-frequency resolution, and cross-terms. To address these issues, this paper presents a new approach called Short-Time Adaptive Compact Kernel Distribution (STACKD). Based on the characteristics of the vibration signals, we designed Chirp-modulated Gaussian Window and used them to segment the overall signal into small segments. Then, we used a two-step Bayesian optimization method to obtain adaptive compact kernel distributions (ACKD) for these small segments of the signal. Finally, we recombined the ACKD of these small segments into a global STACKD. Results from simulations and experiments demonstrate that STACKD offers improved robustness against interference signals, lower computational complexity, and higher time-frequency resolution without cross-term effects. This method is tailored for diagnosing rolling bearings under variable working conditions, providing enhanced visualization of diagnostic processes with superior time-frequency resolution.

A proportional fusion adaptation algorithm to reduce noise in sEMG signals of the lower limb

Abstract To address issues related to unknown noise pollution and the inadequate performance of single adaptive noise reduction algorithms in measuring lower limb surface electromyography (sEMG) signals, this paper proposes an algorithm that combines the fast convergence speed and strong noise reduction capability of recursive least-squares (RLS) with the low computational complexity of the normalized least-mean-square (NLMS) and improved proportionate NLMS (IPNLMS). Through the proportional fusion of RLS with NLMS and RLS with IPNLMS, the proposed algorithm greatly improves convergence speed, stability, and noise reduction performance while effectively reducing computational complexity. Moreover, considering the influence of the initial value of weights on the noise reduction performance during the updating process of the adaptive noise reduction algorithm, a weight initial value setting (WIS) module is proposed to optimize the initial value of the weights by the known amount of data. Based on 50 independent experiments, an adaptive noise reduction algorithm and WIS module were used to reduce the unknown noise in the lower limb sEMG signals, which was generated by white Gaussian noise, power line interference, or hybrid noise interfered by an unknown environment. The noise reduction performance was evaluated by using the average value of the signal-to-noise ratio (SNR), the root-mean-square error, and R-square. Compared with the RLS, NLMS, and IPNLMS algorithms for noise reduction of vastus lateralis, rectus femoris, and biceps femoris signals containing unknown noise, the SNR of RLS-NLMS, RLS-IPNLMS, WIS-RLS-NLMS, and WIS-RLS-IPNLMS is improved by an average of [7.92%, 55.54%, 55.63%], [7.45%, 54.90%, 54.99%], [19.70%, 72.38%, 72.71%], [19.32%, 71.84%, 72.19%]. The simulation results verify that the proportional fusion adaptive noise reduction algorithm and the WIS module effectively accelerate the convergence speed, enhance the noise reduction capability, and reduce the computational complexity.

TCEDN: A Lightweight Time-Context Enhanced Depression Detection Network.

The automatic video recognition of depression is becoming increasingly important in clinical applications. However, traditional depression recognition models still face challenges in practical applications, such as high computational costs, the poor application effectiveness of facial movement features, and spatial feature degradation due to model stitching. To overcome these challenges, this work proposes a lightweight Time-Context Enhanced Depression Detection Network (TCEDN). We first use attention-weighted blocks to aggregate and enhance video frame-level features, easing the model's computational workload. Next, by integrating the temporal and spatial changes of video raw features and facial movement features in a self-learning weight manner, we enhance the precision of depression detection. Finally, a fusion network of 3-Dimensional Convolutional Neural Network (3D-CNN) and Convolutional Long Short-Term Memory Network (ConvLSTM) is constructed to minimize spatial feature loss by avoiding feature flattening and to achieve depression score prediction. Tests on the AVEC2013 and AVEC2014 datasets reveal that our approach yields results on par with state-of-the-art techniques for detecting depression using video analysis. Additionally, our method has significantly lower computational complexity than mainstream methods.

An Adaptive SCG-ECG Multimodal Gating Framework for Cardiac CTA.

Cardiovascular disease (CVD) is the leading cause of death worldwide. Coronary artery disease (CAD), a prevalent form of CVD, is typically assessed using catheter coronary angiography (CCA), an invasive, costly procedure with associated risks. While cardiac computed tomography angiography (CTA) presents a less invasive alternative, it suffers from limited temporal resolution, often resulting in motion artifacts that degrade diagnostic quality. Traditional ECG-based gating methods for CTA inadequately capture cardiac mechanical motion. To address this, we propose a novel multimodal approach that enhances CTA imaging by predicting cardiac quiescent periods using seismocardiogram (SCG) and ECG data, integrated through a weighted fusion (WF) approach and artificial neural networks (ANNs). We developed a regression-based ANN framework (r-ANN WF) designed to improve prediction accuracy and reduce computational complexity, which was compared with a classification-based framework (c-ANN WF), ECG gating, and US data. Our results demonstrate that the r-ANN WF approach improved overall diastolic and systolic cardiac quiescence prediction accuracy by 52.6% compared to ECG-based predictions, using ultrasound (US) as the ground truth, with an average prediction time of 4.83 ms. Comparative evaluations based on reconstructed CTA images show that both r-ANN WF and c-ANN WF offer diagnostic quality comparable to US-based gating, underscoring their clinical potential. Additionally, the lower computational complexity of r-ANN WF makes it suitable for real-time applications. This approach could enhance CTA's diagnostic quality, offering a more accurate and efficient method for CVD diagnosis and management.

Lightweight Detection of Broccoli Heads in Complex Field Environments Based on LBDC-YOLO

Robotically selective broccoli harvesting requires precise lightweight detection models to efficiently detect broccoli heads. Therefore, this study introduces a lightweight and high-precision detection model named LBDC-YOLO (Lightweight Broccoli Detection in Complex Environment—You Look Only Once), based on the improved YOLOv8 (You Look Only Once, Version 8). The model incorporates the Slim-neck design paradigm based on GSConv to reduce computational complexity. Furthermore, Triplet Attention is integrated into the backbone network to capture cross-dimensional interactions between spatial and channel dimensions, enhancing the model’s feature extraction capability under multiple interfering factors. The original neck network structure is replaced with a BiFPN (Bidirectional Feature Pyramid Network), optimizing the cross-layer connection structure, and employing weighted fusion methods for better integration of multi-scale features. The model undergoes training and testing on a dataset constructed in real field conditions, featuring broccoli images under various influencing factors. Experimental results demonstrate that LBDC-YOLO achieves an average detection accuracy of 94.44% for broccoli. Compared to the original YOLOv8n, LBDC-YOLO achieves a 32.1% reduction in computational complexity, a 47.8% decrease in parameters, a 44.4% reduction in model size, and a 0.47 percentage point accuracy improvement. When compared to models such as YOLOv5n, YOLOv5s, and YOLOv7-tiny, LBDC-YOLO exhibits higher detection accuracy and lower computational complexity, presenting clear advantages for broccoli detection tasks in complex field environments. The results of this study provide an accurate and lightweight method for the detection of broccoli heads in complex field environments. This work aims to inspire further research in precision agriculture and to advance knowledge in model-assisted agricultural practices.

Microexpression Recognition Method Based on ADP-DSTN Feature Fusion and Convolutional Block Attention Module

Microexpressions are subtle facial movements that occur within an extremely brief time frame, often revealing suppressed emotions. These expressions hold significant importance across various fields, including security monitoring and human–computer interaction. However, the accuracy of microexpression recognition is severely constrained by the inherent characteristics of these expressions. To address the issue of low detection accuracy regarding the subtle features present in microexpressions’ facial action units, this paper proposes a microexpression action unit detection algorithm, Attention-embedded Dual Path and Shallow Three-stream Networks (ADP-DSTN), that incorporates an attention-embedded dual path and a shallow three-stream network. First, an attention mechanism was embedded after each Bottleneck layer in the foundational Dual Path Networks to extract static features representing subtle texture variations that have significant weights in the action units. Subsequently, a shallow three-stream 3D convolutional neural network was employed to extract optical flow features that were particularly sensitive to temporal and discriminative characteristics specific to microexpression action units. Finally, the acquired static facial feature vectors and optical flow feature vectors were concatenated to form a fused feature vector that encompassed more effective information for recognition. Each facial action unit was then trained individually to address the issue of weak correlations among the facial action units, thereby facilitating the classification of microexpression emotions. The experimental results demonstrated that the proposed method achieved great performance across several microexpression datasets. The unweighted average recall (UAR) values were 80.71%, 89.55%, 44.64%, 80.59%, and 88.32% for the SAMM, CASME II, CAS(ME)3, SMIC, and MEGC2019 datasets, respectively. The unweighted F1 scores (UF1) were 79.32%, 88.30%, 43.03%, 81.12%, and 88.95%, respectively. Furthermore, when compared to the benchmark model, our proposed model achieved better performance with lower computational complexity, characterized by a Floating Point Operations (FLOPs) value of 1087.350 M and a total of 6.356 × 106 model parameters.

HAGMN-UQ: Hyper association graph matching network with uncertainty quantification for coronary artery semantic labeling

Coronary artery disease (CAD) is one of the leading causes of death worldwide. Accurate extraction of individual arterial branches from invasive coronary angiograms (ICA) is critical for CAD diagnosis and detection of stenosis. Generating semantic segmentation for coronary arteries through deep learning-based models presents challenges due to the morphological similarity among different types of coronary arteries, making it difficult to maintain high accuracy while keeping low computational complexity. To address this challenge, we propose an innovative approach using the hyper association graph-matching neural network with uncertainty quantification (HAGMN-UQ) for coronary artery semantic labeling on ICAs. The graph-matching procedure maps the arterial branches between two individual graphs, so that the unlabeled arterial segments are classified by the labeled segments, and the coronary artery semantic labeling is achieved. Leveraging hypergraphs not only extends representation capabilities beyond pairwise relationships, but also improves the robustness and accuracy of the graph matching by enabling the modeling of higher-order associations. In addition, employing the uncertainty quantification to determine the trustworthiness of graph matching reduces the required number of comparisons, so as to accelerate the inference speed. Consequently, our model achieved an accuracy of 0.9211 for coronary artery semantic labeling with a fast inference speed, leading to an effective and efficient prediction in real-time clinical decision-making scenarios.

An illumination-guided dual-domain network for image exposure correction

Exposure problems, including underexposure and overexposure, can significantly degrade image quality. Poorly exposed images often suffer from coupled illumination degradation and detail degradation, aggravating the difficulty of recovery. These necessitate a spatial discriminating exposure correction, making achieving uniformly exposed and visually consistent images challenging. To address these issues, we propose an Illumination-guided Dual-domain Network (IDNet), which employs a Dual-Domain Module (DDM) to simultaneously recover illumination and details from the frequency and spatial domains, respectively. The DDM also integrates a structural re-parameterization technique to enhance the detail-aware capabilities with reduced computational cost. An Illumination Mask Predictor (IMP) is introduced to guide exposure correction by estimating the optimal illumination mask. The comparison with 26 methods on three benchmark datasets shows that IDNet achieves superior performance with fewer parameters and lower computational complexity. These results confirm the effectiveness and efficiency of our approach in enhancing image quality across various exposure scenarios.

Gridless DOA Estimation Method for Arbitrary Array Geometries Based on Complex-Valued Deep Neural Networks

Gridless direction of arrival (DOA) estimation methods have garnered significant attention due to their ability to avoid grid mismatch errors, which can adversely affect the performance of high-resolution DOA estimation algorithms. However, most existing gridless methods are primarily restricted to applications involving uniform linear arrays or sparse linear arrays. In this paper, we derive the relationship between the element-domain covariance matrix and the angular-domain covariance matrix for arbitrary array geometries by expanding the steering vector using a Fourier series. Then, a deep neural network is designed to reconstruct the angular-domain covariance matrix from the sample covariance matrix and the gridless DOA estimation can be obtained by Root-MUSIC. Simulation results on arbitrary array geometries demonstrate that the proposed method outperforms existing methods like MUSIC, SPICE, and SBL in terms of resolution probability and DOA estimation accuracy, especially when the angular separation between targets is small. Additionally, the proposed method does not require any hyperparameter tuning, is robust to varying snapshot numbers, and has a lower computational complexity. Finally, real hydrophone data from the SWellEx-96 ocean experiment validates the effectiveness of the proposed method in practical underwater acoustic environments.

A new approach for neural decoding by inspiring of hyperdimensional computing for implantable intra-cortical BMIs

In the field of Brain Machine Interface (BMI), the process of translating motor intention into a machine command is denoted as decoding. However, despite recent advancements, decoding remains a formidable challenge within BMI. The utilization of current decoding algorithms in the field of BMI often involves computational complexity and requires the use of computers. This is primarily due to the reliance on mathematical models to address the decoding issue and perform subsequent output calculations. Unfortunately, computers are not feasible for implantable BMI systems due to their size and power consumption. To address this predicament, this study proposes a pioneering approach inspired by hyperdimensional computing. This approach first involves identifying the pattern of each stimulus by considering the normal firing rate distribution of each neuron. Subsequently, the newly observed firing pattern for each input is compared with the patterns detected at each moment for each neuron. The algorithm, which shares similarities with hyperdimensional computing, identifies the most similar pattern as the final output. This approach reduces the dependence on mathematical models. The efficacy of this method is assessed through the utilization of an authentic dataset acquired from the Frontal Eye Field (FEF) of two male rhesus monkeys. The output space encompasses eight possible angles. The results demonstrate an accuracy rate of 51.5% while exhibiting significantly low computational complexity, involving a mere 2050 adder operators. Furthermore, the proposed algorithm is implemented on a field-programmable gate array (FPGA) and as an ASIC designe in a standard CMOS 180 nm technology, underscoring its suitability for real-time implantable BMI applications. The implementation required only 2.3 Kbytes of RAM, occupied an area of 2.2 mm2, and consumed 9.32 µW at a 1.8 V power supply. Consequently, the proposed solution represents an accurate, low computational complexity, hardware-friendly, and real-time approach.

A novel highly efficient alternating direction method of multipliers for large-scale trimmed concave SVM

Support vector machine (SVM) is widely utilized for classification in diverse fields thanks to its superior performance. However, when tackling large-scale SVM problems, it encounters high computational complexity. To address this issue, we develop a novel trimmed concave loss SVM model called Ltri-SVM, which achieves sparsity and robustness simultaneously. The new optimality theory for the nonconvex and nonsmooth Ltri-SVM is developed by our new constructed proximal stationary point. Based on which, a new fast alternating direction method of multipliers (ADMM) with working set and low computational complexity is established for addressing Ltri-SVM. Based on our new established optimality theory, our proposed algorithm will divides the training dataset into two distinct categories: the non-working and the working sets. In every training iteration, the algorithm modifies the parameters associated with the working set, while ensuring that the parameters belonging to the non-working set remain unchanged. This algorithm enables the updating of parameters on smaller datasets, which in turn decreases computational complexity and improves the efficiency of runtime. Further, we prove that our algorithm is global convergence. Numerical experiments have confirmed the excellent performance of our algorithm in terms of accuracy in classification, number of support vector and computational speed, surpassing nine other leading solvers. For instance, when solving the real dataset with over 107 samples, our algorithm is able to complete the task in just 18.92 s, which is much faster than other solvers that take at least 650.4 s.

A multimodal multistream multilevel fusion network for finger joint angle estimation with hybrid sEMG and FMG sensing

Finger joint angle (FJA) estimation, as a dynamic and fine-grained decoding mode, can support intuitive and natural human–machine interactions. This study is pioneering work dedicated to achieving accurate and clinically friendly FJA estimation. In this context, a novel surface electromyography and force myography-based multimodal multistream multilevel fusion network (MMMFNet) is proposed. MMMFNet utilises gate-recurrent units (GRU) networks to capture spatial dependencies among features and adopts a multilevel fusion strategy to extract more comprehensive representations from information sources than a single-level fusion. A cross-stream interaction (CSI) block is added between the GRU networks to focus on important information in the prediction and optimise the learning process and representation capability by facilitating information interaction among branch networks. The experimental results suggest that the MMMFNet has higher estimation accuracy (RMSE: 6.225±0.276 vs. 6.953±0.282, R2: 0.879±0.019 vs 0.846±0.021), shorter training time (4.146±0.074 min vs. 16.536±0.430 min), and lower computational complexity (5.4×106 FLOPs vs 8.8×106 FLOPs) than state-of-the-art fusion methods. Moreover, MMMFNet, which quantifies feature importance using attention weights, is the first explainable multimodal model for hand movement intention decoding. In summary, compared with state-of-the-art methods, MMMFNet can not only estimate FJAs more accurately but is also more clinically friendly.

A reservoir computing approach in active noise control for vehicle applications

Reservoir Computing (RC) is an alternative approach to Recurrent Neural Networks (RNN) that can avoid some shortcomings of conventional gradient decent-training that are mainly related to high computational complexity and the reduced ability to "learn" long-range dependencies. In RC the input data are transformed into patterns in a high-dimensional space by an RNN called the "reservoir", which remains unchanged during training. The desired output signal is generated as a linear combination of the neuron's signals from the input-excited reservoir. The linear part of the RC architecture, which is called the "readout" can be trained with a simple learning algorithm such as linear regression. In this work an RC architecture is proposed for the attenuation of acoustic disturbances encountered in vehicle cabins such as aircrafts and yachts. At first, a dataset of acoustic pressure measurements from real-world cabins was used to generate the readout and reservoir layers. Subsequent computer simulations demonstrated that the proposed architecture can successfully attenuate such acoustic disturbances. Finally, the performance of RC in Active Noise Cancellation task was compared with both the linear FxLMS algorithm and a conventional RNN, showing that it has several advantages regarding acoustic pressure reduction combined with relatively low computational complexity.