Neural Network Filter Research Articles

Suppression of interference fringes in wavelength modulation spectroscopy based on convolutional neural networks

To mitigate interference fringes in wavelength modulated spectroscopy (WMS), we introduce a new filter based on convolutional neural networks (New CNNF). This filter effectively suppresses both high and low-frequency variations of interference fringes, including those with envelopes. In striving for comprehensive interference fringe suppression across the entire frequency spectrum, while accommodating diverse cavity length conditions, we confront the challenge of data sparsity. To address this, we introduce a pioneering methodology: finely segmenting the dataset through concentration column density normalization, thereby achieving notable noise reduction. Through the construction and training of the New CNNF model, it was found to exhibit superior performance compared to traditional filtering algorithms, particularly under low signal-to-noise ratio (SNR) conditions.Following processing with the novel convolutional neural network filters (New CNNF), the sample's signal-to-noise ratio (SNR) improved by 26.70 dB, increasing from the original −10.11 dB. When the etalon lengths were 2 cm and 100 cm, the goodness of fit between the predicted second harmonic signal amplitude by New CNNF and the corresponding label concentration reached 0.9995 and 0.9999, respectively. Experimental results demonstrate that New CNNF effectively suppresses high and low-frequency variations and enveloped interference fringes in the TDLAS-WMS system, thereby enhancing the accuracy and stability of methane concentration measurements. CH4 transitions at λ = 1.654 μm were selected to validate this approach. Our method shows promising application prospects and can be extended to the sensing of other gas molecules.

Non-invasive brain-machine interface control with artificial intelligence copilots.

Motor brain-machine interfaces (BMIs) decode neural signals to help people with paralysis move and communicate. Even with important advances in the last two decades, BMIs face key obstacles to clinical viability. Invasive BMIs achieve proficient cursor and robotic arm control but require neurosurgery, posing significant risk to patients. Non-invasive BMIs do not have neurosurgical risk, but achieve lower performance, sometimes being prohibitively frustrating to use and preventing widespread adoption. We take a step toward breaking this performance-risk tradeoff by building performant non-invasive BMIs. The critical limitation that bounds decoder performance in non-invasive BMIs is their poor neural signal-to-noise ratio. To overcome this, we contribute (1) a novel EEG decoding approach and (2) artificial intelligence (AI) copilots that infer task goals and aid action completion. We demonstrate that with this "AI-BMI," in tandem with a new adaptive decoding approach using a convolutional neural network (CNN) and ReFIT-like Kalman filter (KF), healthy users and a paralyzed participant can autonomously and proficiently control computer cursors and robotic arms. Using an AI copilot improves goal acquisition speed by up to 4.3× in the standard center-out 8 cursor control task and enables users to control a robotic arm to perform the sequential pick-and-place task, moving 4 randomly placed blocks to 4 randomly chosen locations. As AI copilots improve, this approach may result in clinically viable non-invasive AI-BMIs.

Dynamic data reconciliation based on elman neural network and particle filter

In the process of modern industries, complex nonlinear dynamic systems present high requirements for measured data. In the actual industrial process system, the measurement data obtained by sensors will inevitably be subject to noise disturbances from the equipment itself or from the outside environment. These noise disturbances will deteriorate the dynamic performance of the system to a certain extent and affect the industrial production. Particle filter (PF) can be used to infer the accurate outputs of nonlinear dynamic system from the contaminated measurement data, but PF is limited to the pre-known state space model. In the actual industrial process, it is difficult to summarize the internal behavior of the system and obtain the pre-known state space model. Therefore, it is impossible to directly use PF in the nonlinear dynamic system with unknown model. In order to solve the above problems, this paper proposes a dynamic data reconciliation method called ENN-PF, which combines Elman neural network (ENN) data-driven modeling with PF. In this method, ENN is used for data-driven modeling, that is, the system model is dynamically identified by using the input and output data of the system, and then the dynamic data reconciliation is carried out by using PF according to the model identified by ENN. Finally, the proposed ENN-PF method is validated by simulations and practical experiments to effectively reduce the interference of measurement noise and improve the dynamic performance of the system.

Enhancing physically-based flood forecasts through fusion of long short-term memory neural network with unscented Kalman filter

Accurate flood forecasts are vital for reservoir operation and flood prevention. The unscented Kalman filter (UKF) excels in improving physically-based models flood forecasting but depends on precise noise estimation, posing challenges in complex uncertainty quantification. In this study, we propose a novel approach, the deep fusion of a long short-term memory (LSTM) neural network and unscented Kalman filter (LSTM-UKF), to enhance flood forecasts by adaptively updating hydrological model states. The LSTM efficiently learns noise-related information from the estimation of Kalman Gain, obviating the need for intricate uncertainty recognition and approximation. For comparative analysis, the LSTM-UKF and UKF filters are constructed to correct Xinanjiang (XAJ) hydrological model states. Both filters utilize streamflow observations to update the XAJ model states and facilitate its flood forecasting performance. Comprehensive evaluations were conducted using 3-hourly meteor-hydrological sequences from the Jianxi and Tianyi basins in China, focusing on the accuracy, stability, and reliability of multi-step-ahead flood forecasts. Results indicate that the LSTM-UKF performs superior to the UKF, with maximal advancements in Nash-Sutcliffe efficiency coefficient (NSE) by 9.1%, maximal reduction in root mean square error (RMSE) by 18.7%, and maximal decrease in mean absolute error (MAE) by 22.6%. Additionally, the LSTM-UKF exhibits better robustness and stability in non-stationary flood events. The proposed LSTM-UKF mitigates the over-reliance on noise estimates, reducing systematic biases and error accumulation in state estimates and enhancing hydrological model generalizability in operational flood prevention platforms and systems.

Mid-infrared methane standoff sensor using a frequency channel attention based convolutional neural network filter

Highly sensitive standoff methane detection is vital for atmospheric science, environmental protection, and production safety. We develop a mid-infrared methane standoff sensor using a cooperative target based on tunable diode laser absorption spectroscopy (TDLAS). To enhance sensor sensitivity, we propose a one-dimensional frequency channel attention-based convolutional neural network (1D-FCACNN) filter, which can effectively denoise the methane absorbance signals and its second harmonic signals. The filter model is adequately trained on a simulated spectrum dataset, which we constructed based on data augmentation. In comparison with reported filtering algorithms, the proposed filter shows the best performance in both measuring modes and evaluation metrics. Real-time measurements show that the measuring accuracy and limit of detection (LOD) of the proposed sensor reach a minimum of 30.40 ppb and 5.04 ppb over a 10-meter optical range, a significant improvement compared to previous reports of methane standoff sensors. The proposed methane standoff sensor proves the feasibility of enhancing the performance of TDLAS gas sensors with the attention mechanism, bringing a new option for high-sensitivity measurements of methane and other atmospheric trace gases.

Fusing LSTM neural network and expanded disturbance Kalman filter for estimating external disturbing forces of ball screw drives

Monitoring external disturbing forces is of great significance for improving the control performance and interaction safety of ball screw drives. In consideration of the low cost in long-term use and the non-invasiveness to work space, estimating external disturbing forces using motor torque and motion states has been viewed as the solution that has great potential to be applied in real industry. However, the existing methods cannot efficiently establish accurate estimation models in a low-cost and general manner (i.e., not depend on specific work scenarios). To overcome this challenge, a novel hybrid-driven method that fuses Long Short-Term Memory (LSTM) Neural Network (NN) and Expanded Disturbance Kalman Filter (EDKF) is proposed. The motion state of the worktable is decoupled as the linear superposition of the outputs of two independent dynamic models, i.e., Ideal Dynamic Model (IDM) and Disturbance Dynamic Model (DDM). IDM predicts the motion state that is only driven by the motor torque, is modeled as an LSTM NN and trained by self-excitation experiments. DDM is equivalent to the disturbance transfer function and easily identified by impact tests. The motion state only driven by external disturbing forces is calculated by the predicted result of IDM and the monitored actual motion state, and then external disturbing forces are estimated by a DDM-based EDKF. The verifications in simulation and real environments both show that by the proposed method, accurate estimation models of external disturbing forces can be efficiently established through self-excitation experiments and impact tests.

State-of-charge estimation of lithium-ion battery: Joint long short-term memory network and adaptive extended Kalman filter online estimation algorithm

With the widespread use of lithium-ion batteries, estimating the State of Charge (SOC) has been one of the most critical tasks in the battery management system (BMS). In this paper, a joint long short-term memory recurrent neural network (LSTM-RNN) and adaptive extended Kalman filter (AEKF) algorithm is proposed to achieve accurate SOC estimation. The proposed algorithm generates an initial SOC estimation value through a trained LSTM-RNN with a sliding time window signal sequence as input. The initial estimation value is then used as a feedback input to the AEKF. The results show that only a 30-s time series as the input of the LSTM-RNN can achieve satisfactory estimation results. Compared with traditional adaptive extended Kalman filter and unscented Kalman filter (UKF), the proposed algorithm has a better performance of estimation accuracy under various discharging conditions, especially when applied at low temperatures. Finally, the SOC estimation results under four operating conditions and various temperatures show that the maximum error is less than 2.5 %, in particular, the root mean square error (RMSE) and the mean absolute error (MAE) are below 1.01 % and 0.64 %. The proposed algorithm is an effective SOC estimation method with good robustness and high estimation accuracy.

State estimation of buses: A hybrid algorithm of deep neural network and unscented Kalman filter considering mass identification

Accurately obtaining vehicle state parameters is a crucial prerequisite for active safety control. However, key stability parameters of buses, such as sideslip angle and roll angle, are difficult to measure directly. Moreover, changes in the bus mass during the driving route can impact the vehicle state. To tackle these challenges, we identify the bus mass before state estimation, and propose a hybrid state estimation algorithm based on deep neural network (DNN) and unscented Kalman filter (UKF). First, the mass is identified using the recursive least squares (RLS) method based on longitudinal dynamics at each start of bus, and the yaw and roll moment of inertia are adaptively adjusted based on the identified mass. Second, a composite DNN combining the convolutional neural network (CNN), gated recurrent unit (GRU), and attention mechanism is designed to estimate sideslip and roll angles. As for the training dataset, it is acquired from different maneuvers simulated by TruckSim. Third, a UKF estimator is established based on 4-degree of freedom (DOF) vehicle dynamics model and magic tire formula, and the estimated values of DNN are inserted into UKF estimator as virtual observations. Finally, the proposed hybrid algorithm is validated through simulation maneuvers and real driving maneuvers based on MATLAB/Simulink and TruckSim software. The comparative results demonstrate that the proposed algorithm outperforms individual DNN estimation and UKF estimation, enhancing both estimation accuracy and reliability.

Highly sensitive mid-infrared methane remote sensor using a deep neural network filter.

A novel mid-infrared methane remote sensor integrated on a movable platform based on a 3.291-µm interband cascade laser (ICL) and wavelength modulation spectroscopy (WMS) is proposed. A transmitting-receiving coaxial, visualized optical layout is employed to minimize laser energy loss. Using a hollow retro-reflector remotely deployed as a cooperative target, the atmospheric average methane concentration over a 100-meter optical range is measured with high sensitivity. A deep neural network (DNN) filter is used for second harmonic (2f) signal denoising to compensate for the performance shortcomings of conventional filtering. Allan deviation analysis indicated that after applying the DNN filter, the limit of detection (LOD) of methane was 86.62 ppb with an average time of 1 s, decreasing to 12.03 ppb with an average time of 229 s, which is a significant promotion compared to similar work reported. The high sensitivity and stability of the proposed sensor are shown through a 24-hour continuous monitoring experiment of atmospheric methane conducted outdoors, providing a new solution for high-sensitivity remote sensing of atmospheric methane.

Indirect Adaptive Control Using Neural Network and Discrete Extended Kalman Filter for Wheeled Mobile Robot

This paper presents a novel approach to address the challenges associated with the trajectory tracking control of wheeled mobile robots (WMRs). The proposed control approach is based on an indirect adaptive control PID using a neural network and discrete extended Kalman filter (IAPIDNN-DEKF). The proposed IAPIDNN-DEKF scheme uses the NN to identify the system Jacobian, which is used for tuning the PID gains using the stochastic gradient descent algorithm (SGD). The DEKF is proposed for state estimation (localization), and the NN adaptation improves the tracking error performance. By augmenting the state vector, the NN captures higher-order dynamics, enabling more accurate estimations, which improves trajectory tracking. Simulation studies in which a WMR is used in different scenarios are conducted to evaluate the effectiveness of the IAPIDNN-DEKF control. In order to demonstrate the effectiveness of the IAPIDNN-DEKF control, its performance is compared with direct adaptive NN (DA-NN) control, backstepping control (BSC) and an adaptive PID. On lemniscate, IAPIDNN-DEKF achieves RMSE values of 0.078769, 0.12086 and 0.1672. On sinusoidal trajectories, the method yields RMSE values of 0.01233, 0.015138 and 0.088707, and on sinusoidal with perturbation, RMSE values are 0.021495, 0.016504 and 0.090142 in x, y and θ, respectively. These results demonstrate the superior performance of IAPIDNN-DEKF for achieving accurate control and state estimation. The proposed IAPIDNN-DEKF offers advantages in terms of accurate estimation, adaptability to dynamic environments and computational efficiency. This research contributes to the advancement of robust control techniques for WMRs and showcases the potential of IAPIDNN-DEKF to enhance trajectory tracking and state estimation capabilities in real-world applications.

A Model for Detecting the Presence of Pesticide Residues in Edible Parts of Tomatoes, Cabbages, Carrots, and Green Pepper Vegetables

With increased resistant pests and low crop yields, farmers especially in sub-Saharan Africa have greatly embraced usage of chemicals. These chemicals include pesticides used in gardens for better yields and also in the stalls for longer shelf life by sellers of farm products especially fresh perishables like tomatoes, cabbages, carrots, and green pepper vegetables. This, if not checked, may expose humans and animals to pesticide residues. In this research, a model for detecting the presence of pesticide residues in edible parts of vegetables (tomatoes, cabbages, carrots, and green pepper) was developed. A dataset consisting of 1094 images of both contaminated and uncontaminated vegetables including tomatoes, cabbages, carrots, and green pepper with a scale magnification of 800 × 1276 pixels taken using InfiRay P2 pro Night Vision Go Mini Infrared Thermal camera with a thermal module was taken from different daily markets in Mbarara city, South Western Uganda. Image preprocessing was done by noise removal and grayscale conversion. Both the neural network and median filter were applied on the images. A python script was used to cluster the dataset based on chemical concentrations rates of 0.1–0.8 mg/kg, 0.9–1.3 mg/kg, and 1.4–1.7 mg/kg, and this was done for both training and testing dataset. Feature extraction was done to detect the presence of mancozeb, dioxacarb, and methidathion residues from the cleaned images. To test the developed model, convolutional neural networks transfer learning models, Inception V3, VGG16, VGG19, ResNet50, and the scratch model were used. From the results obtained, Inception V3 achieved better performance compared to other transfer learning models with 96.77% followed by VGG16 at 86.98%, VGG19 at 87.56%, and ResNet50 at 82.11%, whereas the developed scratch model achieved 89.13% classification accuracy.

Freedom of choice and network "echo effects" of information consumption

The mechanisms of media personalization and separation in network communication on the principle of like-mindedness stimulate the formation of "echo chambers" and "filter bubbles". The social network phenomenon indicated by these metaphors is the subject of research, and the concepts themselves are considered as close, but not identical. The purpose of the study is to identify factors affecting the effectiveness of filtering algorithms, the user's ability to make informed choices and self–organization in the process of information consumption. The analysis of the causes and consequences of selective strategies of online information consumption is carried out on the basis of the communicative - activity approach, the theory of virtual reality Zh. Baudrillard, concepts of digital media by K. Sunstein, E. Pariser and R. Fletcher. The results of foreign and domestic studies of communication practices on various social platforms are used as an empirical basis. The research approaches presented in the scientific literature allow us to focus on the technological and logical-semantic perspective of the analysis of stable forms of network communication. According to the authors, the interdependence of filtering algorithms and value dominants of information consumption leaves the user with a chance to independently choose a "consumer basket". The duality in obtaining personalized content is emphasized: on the one hand, it is convenience, saving effort, on the other - the one-dimensionality of the world picture in the information bubble. Based on this, freedom of choice is characterized as the right to active choice and the right not to choose, consciously delegating it to neural network filters. In conclusion, the authors identified internal and external network factors for reducing the effectiveness of filtering algorithms: the interpretation of user behavior by artificial intelligence; the functioning of rational confrontational communication; the opportunistic conceptualization of echo effects; the availability of means of conscious counteraction. Incentives for reasonable information consumption, technological and cognitive ways of protecting and rationalizing user behavior are highlighted.

Neural Network-Based Approach to Detect and Filter Misleading Audio Segments in Classroom Automatic Transcription

Audio recording in classrooms is a common practice in educational research, with applications ranging from detecting classroom activities to analyzing student behavior. Previous research has employed neural networks for classroom activity detection and speaker role identification. However, these recordings are often affected by background noise that can hinder further analysis, and the literature has only sought to identify noise with general filters and not specifically designed for classrooms. Although the use of high-end microphones and environmental monitoring can mitigate this problem, these solutions can be costly and potentially disruptive to the natural classroom environment. In this context, we propose the development of a novel neural network model that specifically detects and filters out problematic audio sections in classroom recordings. This model is particularly effective in reducing transcription errors, achieving up to a 96% success rate in filtering out segments that could lead to incorrect automated transcriptions. The novelty of our work lies in its targeted approach for low-budget, aurally complex environments like classrooms, where multiple speakers are present. By allowing the use of lower-quality recordings without compromising analysis capability, our model facilitates data collection in natural educational settings and reduces the dependency on expensive recording equipment. This advancement not only demonstrates the practical application of specialized neural network filters in challenging acoustic environments but also opens new avenues for enhancing audio analysis in educational research and beyond.

Study of non-linear filter-based algorithms for active noise control of machinary noises

Sometimes, machinary noise comprises of non-linear deterministic noise. Therefore, it is expected that for such cases of noises, non-linear filter based active noise control algorithms can provide improved performance in terms of reduction of noise. Various non-linear filter based algorithms explored in this paper are: Volterra filtered-x least mean square algorithm, Hammerstein algorithm, Fourier algorithm utilising Fourier non-linear filter, even mirror Fourier non-linear filter and CN filter, bilinear algorithm, Legendre algorithm, filtered-s least mean square using functional link artifical neural network (FLANN) filter, generalised FLANN, exponential FLANN and recursive FLANN. Noise from various workshop machines like band saw, compressor and welding process were recorded using a microphone and utilised to study the algorithms. Two cases of primary paths, (a) linear primary path and (b) non-linear primary path with cubic non-linearity, are considered for simulation study. Performance of various algorithms is evaluated based on steady-state residual noise.

A neural network filter based high-sensitive MIR CO2 sensor

A low-power consumption, small-size and high measurement sensitivity mid-infrared (MIR) carbon dioxide (CO2) sensor was developed based on wavelength modulation spectroscopy (WMS) technology, utilizing a mid-infrared distributed feedback interband cascade laser (DFB-ICL) operating around 2682 nm. The sensor incorporated two key components, a home-made multi-pass gas cell with base length of 10 cm achieving 1.5 m optical path and a highly integrated FPGA board, which were used to reduce the size of the sensor. A novel second harmonic-neural network filter (SH-NNF) was also proposed and implemented in this sensor. The sensor’s performance was evaluated through a 20 mins continuous experiment using calibrated gas with 500 ppm concentration. The results, before and after SH-NNF, demonstrated an improvement in sensitivity and detection precision from 1.68 ppm (1 σ) to 1.17 ppm and from 2.14 ppm to 1.45 ppm, respectively. The optimum limit of detection of 0.392 ppm with integration time of 218 s was obtained. The ppm detection and compact characteristics of the developed MIR sensor show its potential for detecting CO2 in the ambient air.

State of charge estimation for energy storage lithium-ion batteries based on gated recurrent unit neural network and adaptive Savitzky-Golay filter

State of charge estimation for energy storage lithium-ion batteries based on gated recurrent unit neural network and adaptive Savitzky-Golay filter

Vehicle Sideslip Angle Estimation Based on Radial Basis Neural Network and Unscented Kalman Filter Algorithm

Most existing ESC (electronic stability control) and ADS (auto drive system) stability controls rely on the measurement of yaw rate and sideslip angle. However, the existing sensors are too expensive, which is one of the factors that makes it difficult to measure the side slip angle of vehicles directly. Therefore, the estimation of sideslip angle has been extensively discussed in the relevant literature. Accurate modeling is complicated by the fact that vehicles are highly nonlinear. This article combines a radial basis function neural network with an unscented Kalman filter to propose a new sideslip angle estimation method for controlling the dynamic behavior of vehicles. Considering the influence of input data type and sensor ease of measurement factors on the results, a two-degrees-of-freedom vehicle nonlinear dynamic model was established, and a radial basis function neural network estimation algorithm was designed. In order to reduce the impact of noise and improve the reliability of the algorithm, the neural network algorithm was combined with the Kalman filter. The information collected from low-cost sensors for actual vehicle operation (longitudinal vehicle speed, steering wheel angle, yaw rate, lateral acceleration) was trained using a radial basis function neural network to obtain a “pseudo slip angle”. The “pseudo slip angle”, yaw rate, and lateral acceleration are input as observations of the Kalman filter. The sideslip angle obtained from different observation methods was compared with the values provided by the Carsim 2020. The experiment shows that the sideslip angle estimator based on the radial basis function neural network and unscented Kalman filter achieves the optimal effect.

IIRI-Net: An interpretable convolutional front-end inspired by IIR filters for speaker identification

Learning interpretable filters in Convolutional Neural Networks (CNNs) is an approach that helps to build models with better generalization ability. Interpretable filters can reveal some hidden aspects of the task and help to improve the model. One of the most successful approaches in the field of the speech processing is SincNet, where the model learns some band-pass filters in the first layer of a CNN with a raw waveform as its input. In this paper, similar to SincNet, some meaningful filters are proposed, which here are inspired by Infinite Impulse Response (IIR) filters. The proposed model uses a phase correction process to ensure that phase linearity is satisfied. The effective length of the truncated IIR filter is calculated based on the accumulated energy, and the effect of changing the filter size on the final results has been investigated. The proposed model is evaluated in the speaker identification task on the TIMIT and Librispeech datasets and compared with traditional CNNs and four interpretable kernel-based models. The experimental results show the superiority of the proposed model both in performance and convergence speed. Moreover, some patterns of the speech signal, which lead to uniquely identifying a speaker, are analyzed by examining the spectrum of the learned filters.

Enhancement of Product-Inspection Accuracy Using Convolutional Neural Network and Laplacian Filter to Automate Industrial Manufacturing Processes

The automation of the manufacturing process of printed circuit boards (PCBs) requires accurate PCB inspections, which in turn require clear images that accurately represent the product PCBs. However, if low-quality images are captured during the involved image-capturing process, accurate PCB inspections cannot be guaranteed. Therefore, this study proposes a method to effectively detect defective images for PCB inspection. This method involves using a convolutional neural network (CNN) and a Laplacian filter to achieve a higher accuracy of the classification of the obtained images as normal and defective images than that obtained using existing methods, with the results showing an improvement of 11.87%. Notably, the classification accuracy obtained using both a CNN and Laplacian filter is higher than that obtained using only CNNs. Furthermore, applying the proposed method to images of computer components other than PCBs results in a 5.2% increase in classification accuracy compared with only using CNNs.

A collaborative recursive generalized exponential functional link artificial neural network nonlinear filter based on optional parameter entry method

Different types of exponential functional link artificial neural network (EFLN) filters have been proposed for nonlinear active noise control (NANC) systems, but they still suffer from high computational complexity. To address this issue and further improve steady-state performance, this paper presents a collaborative recursive generalized exponential functional link artificial neural network nonlinear filter based on optional parameter entry method (CR-GEFLN-OPEM). A collaborative filtered-error least mean square (CFELMS) algorithm is also proposed. Simulations and an actual noise control trial demonstrate the improved steady-state performance and low computational complexity of the proposed filter in NANC systems.