Time-frequency Feature Extraction Research Articles

Intelligent fault diagnosis based on multi-sensor data fusion and multi-scale dual attention enhanced network

Abstract As a significant component of rotating machinery, the health of rolling bearings and gears directly impact the normal operation of rotating machinery. Traditional single-sensor fault diagnosis approaches often fail to extract sufficient fault information, resulting in low diagnostic accuracy in practical engineering applications. Additionally, conventional multi-sensor fusion diagnosis methods exhibit the low robustness in noisy environments. To tackle these challenges, this paper presents a novel fault diagnosis approach derived from multi-sensor data fusion and a multi-scale dual attention enhanced network (MSDF-MSDAENet). Initially, a multi-sensor data fusion (MSDF) strategy is presented. This method reduces the dimensionality of high-dimensional data from multiple sensors to extract embedded low-dimensional effective information, which is then fused into a three-dimensional pixel matrix image. Subsequently, a multi-scale dual attention enhanced network (MSDAENet) is constructed, incorporating both time-frequency information feature extraction modules and position information extraction modules. These enhancements significantly enhance the model's capacity to extract fault features. By dividing the RGB fused images into a well-organized dataset and inputting it into the network for fault diagnosis, the method achieves intelligent diagnosis of samples with different fault types. Tests were carried out on three datasets, and the presented method was compared with various existing methods. The ultimate experimental results demonstrate the efficacy and superiority of the presented method.

Artificial intelligence models using F-wave responses predict amyotrophic lateral sclerosis.

Nerve conduction F-wave studies contain critical information about subclinical motor dysfunction which may be used to diagnose patients with amyotrophic lateral sclerosis (ALS). However, F-wave responses are highly variable in morphology, making waveform interpretation challenging. Artificial Intelligence techniques can extract time-frequency features to provide new insights into ALS diagnosis and prognosis. A retrospective analysis was performed on F-wave responses from 46,802 patients. Discrete wavelet transforms were applied to time-series waveform responses after stimulating ulnar, median, fibular, and tibial nerves. Wavelet coefficient statistics, onset age, sex, and BMI were features for training a Gradient Boosting Machine model on 40,095 (5,329 diagnosed with motor neuron disease). Model performance was tested on responses from 689 ALS patients meeting Gold Coast criteria and 689 age- and sex-matched controls. An exploratory analysis examined model performance on cohorts of patients with inclusion body myositis (IBM), cervical radiculopathy, lumbar radiculopathy, or peripheral neuropathy which can mimic ALS symptoms. Factors affecting survival were estimated through cox proportional hazards regression. The model trained using wavelet-features on the full waveform had 90% recall, 87% precision, and 88% accuracy. Similar model performance was measured using features only from the M-Wave or F-Wave. Classification probabilities for ALS patients were statistically different from the diagnoses mimicking ALS symptoms (p<0.001, ANOVA, Tukey's post-hoc), Higher model classification probabilities of ALS, older age at onset, and family history of ALS alone or with frontotemporal dementia were factors decreasing survival. Longer diagnostic delay and upper limb onset site were factors increasing survival. Model scores two standard deviations below the mean had 4 months increased survival (two standard deviations below had 3 months decreased survival). Artificial intelligence techniques extracted important information from F-wave responses to estimate a patient's likelihood of ALS and their survival risks. Although the model can make predictions at specific decision threshold as presented here, the true strength of such a model lies in its ability to provide probabilities about whether a patient is likely to have ALS compared to other mimicking diagnoses such as IBM, cervical or lumbar radiculopathy, or peripheral neuropathy. These probabilities provide clinicians with additional information they can use to make the final diagnosis with greater confidence and precision. Integrating such a model into the clinical workflow could help clinicians diagnose ALS sooner and manage treatment based on estimated survival, which may improve outcomes and patients' quality of life.

A Feature Fusion Model Based on Temporal Convolutional Network for Automatic Sleep Staging Using Single-Channel EEG.

Sleep staging is a crucial task in sleep monitoring and diagnosis, but clinical sleep staging is both time-consuming and subjective. In this study, we proposed a novel deep learning algorithm named feature fusion temporal convolutional network (FFTCN) for automatic sleep staging using single-channel EEG data. This algorithm employed a one-dimensional convolutional neural network (1D-CNN) to extract temporal features from raw EEG, and a two-dimensional CNN (2D-CNN) to extract time-frequency features from spectrograms generated through continuous wavelet transform (CWT) at the epoch level. These features were subsequently fused and further fed into a temporal convolutional network (TCN) to classify sleep stages at the sequence level. Moreover, a two-step training strategy was used to enhance the model's performance on an imbalanced dataset. Our proposed method exhibits superior performance in the 5-class classification task for healthy subjects, as evaluated on the SHHS-1, Sleep-EDF-153, and ISRUC-S1 datasets. This work provided a straightforward and promising method for improving the accuracy of automatic sleep staging using only single-channel EEG, and the proposed method exhibited great potential for future applications in professional sleep monitoring, which could effectively alleviate the workload of sleep technicians.

Real-time in-situ coatings corrosion monitoring using machine learning-enhanced triboelectric nanogenerator

Current methods for monitoring coating corrosion are limited by their inability to provide real-time data and dependence on external power sources. This study presents a novel in-situ corrosion monitoring system using a solid-liquid triboelectric nanogenerator (TENG) that converts mechanical energy into electrical signals for self-powered sensing. TENG signals and electrochemical impedance spectra were measured on a dopamine-modified lignin-polydimethylsiloxane coating on steel in 1 M NaCl solution under no corrosion, indentation, pitting, and broken conditions, respectively. We extract time-frequency features from the TENG signals to predict the coating’s corrosion condition by applying a customised convolutional neural network (CNN). By extracting time-frequency features from the TENG signals and applying a custom CNN, a prediction accuracy of 99 % for corrosion classification was achieved. Furthermore, the CNN regression model predicted coating impedance values with a high coefficient of determination (R² = 0.98), demonstrating its effectiveness in tracking corrosion progression. The developed TENG also facilitates defect localisation via a matrix electrode beneath the coating. Our approach introduces a promising real-time technology for in-situ corrosion monitoring.

Epileptic seizure prediction via multidimensional transformer and recurrent neural network fusion

BackgroundEpilepsy is a prevalent neurological disorder in which seizures cause recurrent episodes of unconsciousness or muscle convulsions, seriously affecting the patient’s work, quality of life, and health and safety. Timely prediction of seizures is critical for patients to take appropriate therapeutic measures. Accurate prediction of seizures remains a challenge due to the complex and variable nature of EEG signals. The study proposes an epileptic seizure model based on a multidimensional Transformer with recurrent neural network(LSTM-GRU) fusion for seizure classification of EEG signals.MethodologyFirstly, a short-time Fourier transform was employed in the extraction of time-frequency features from EEG signals. Second, the extracted time-frequency features are learned using the Multidimensional Transformer model. Then, LSTM and GRU are then used for further learning of the time and frequency characteristics of the EEG signals. Next, the output features of LSTM and GRU are spliced and categorized using the gating mechanism. Subsequently, seizure prediction is conducted.ResultsThe model was tested on two datasets: the Bonn EEG dataset and the CHB-MIT dataset. On the CHB-MIT dataset, the average sensitivity and average specificity of the model were 98.24% and 97.27%, respectively. On the Bonn dataset, the model obtained about 99% and about 98% accuracy on the binary classification task and the tertiary upper classification task, respectively.ConclusionThe findings of the experimental investigation demonstrate that our model is capable of exploiting the temporal and frequency characteristics present within EEG signals.

Harmonizing Insights: LR-FFT Feature Extraction for Alzheimer’s and Autism Spectrum Disorder Detection in EEG Signals

Alzheimer’s Disease (AD) and Autism Spectrum Disorder (ASD) represent two distinct but equally impactful challenges in the field of neurology and cognitive health. AD is a degenerative neurological condition characterized by its progressive nature, typically affecting individuals in later stages of life. The hallmark features include cognitive impairment, mem- ory deterioration, and alterations in behavior. In contrast, ASD is a developmental disorder typically diagnosed in childhood, marked by difficulties in social interaction, communication, and repetitive behaviors. This paper explores the potential of the time-frequency feature extraction model known as the Left-Right Fast Fourier Transform (LR-FFT) in the context of these two disorders. While AD and ASD differ significantly in their onset, presentation, and demographic affected, both necessitate early and accurate diagnosis to enable timely intervention and tailored treatment strategies. Our research yields promising results, with classification accuracies reaching 88.26% for AD and 99.86% for ASD, demonstrating the LR-FFT’s potential to enhance diagnostic accuracy. By contributing to improved differentiation between these complex neurological conditions, this work aims to advance our understanding and management of AD and ASD, ultimately benefiting patients, their families, and healthcare practitioners.

A multi-dimensional machine learning framework for accurate and efficient battery state of charge estimation

Accurate state of charge (SOC) estimation is essential for battery safe and efficient utilization. As artificial intelligence technologies evolve, data-driven methods have become mainstream for estimating SOC. However, the technique can significantly deteriorate model performance when encountering poor or insufficient data quality. In this paper, we apply median filtering to eliminate extreme noise and utilize continuous wavelet transform to extract time-frequency features from voltage signals. Additionally, we generate novel features via feature crossing. We then apply dimensionality reduction via the random forest method to decrease computational expense. Finally, we select a convolutional neural network (CNN) as the base model to learn optimized features for more precise SOC estimation. To confirm the efficacy of our proposed method, this study compares it with CNN, long short-term memory (LSTM), bidirectional LSTM (BILSTM), and a CNN-BILSTM model combined with an attention mechanism. These comparisons are conducted under different temperatures and operating conditions. The results indicate that this method achieves a mean absolute error and a root mean square error of less than 2.89 % and 3.71 %, respectively, in SOC estimation, demonstrating superior accuracy compared to other models. This study underscores the significance of feature engineering techniques in SOC estimation.

Enhancing acceleration data cleaning through time–frequency feature integration using deep learning

In recent years, the safety issues of bridge has garnered significant attention from various sectors of society. Bridge structural health monitoring (SHM) systems have emerged as essential tools for real-time monitoring and assessing the operational safety of bridges. Despite the substantial implementation and operation of monitoring systems, challenges have arisen from abnormal data caused by equipment malfunctions, power outages, and environmental interference. These anomalies are challenging to manually remove within the vast dataset. To address the challenge of automated detection and classification of anomalies, this paper focuses on acceleration data and proposes a data cleaning method that integrates time–frequency features and one-dimensional convolutional neural network (1D CNN). The method utilizes extreme value envelopes and wavelet scattering transforms to extract time–frequency features from acceleration data, thereby achieving data dimensionality reduction and feature extraction goals. Subsequently, these features are employed to train and test a 1D CNN. The proposed method’s validation is conducted using acceleration monitoring data obtained from a large-span arch bridge. Through extensive testing, the model demonstrates a classification accuracy exceeding 98% on both the test set and unseen data, showcasing high classification precision and generalization capabilities. This approach provides an effective solution for automated data cleaning and anomaly identification within the context of bridge health monitoring systems.

New oscillatory features of electrodermal activity signals evaluated in automated mental workload monitoring

Oscillatory components are discriminative features of mental workload monitoring in electrodermal activity (EDA) signals. Currently, most feature extraction techniques mainly focus on time domain or linear techniques but rarely consider the non-stationary nature, individual differences, and recording noise. This study has proposed a new time–frequency feature extraction method based on Tunable Q-factor Wavelet Transform providing an effective quantification of oscillatory behavior of a single transient. The extracted features have been classified using a recurrent neural network with the advantage of dynamic mapping procedure. The proposed method has been evaluated using EDA data acquired during an arithmetic task of five workload levels with both subtle and distinct differences. In a comparative study, the effects of several decomposition parameters and cognitive factors have been explored. Experimental results have shown that it achieves an average accuracy rate of 98.52% for three workload levels discrimination by optimizing the number of features to 3 (features extracted from a low frequency sub-band) from 15 (features extracted from five frequency sub-bands), i.e., 80% feature reduction. The greatest contribution coming fromthe smallest frequency sub-bands has also been obtained. This method is superior to other existing methods in separating 3 to 5 workload levels with both distinct and subtle differences extracting effective oscillatory characteristics. Considering the merits of a simple feature extraction procedure and cost-effectiveness of a single lead EDA signal as well as high discriminability between several workload levels, it provides a trade-off between performance and computational complexity, thus demonstrating its potential for online human-error prevention systems.

Exploring Feature Selection and Classification Techniques to Improve the Performance of an Electroencephalography-Based Motor Imagery Brain-Computer Interface System.

The accuracy of classifying motor imagery (MI) activities is a significant challenge when using brain-computer interfaces (BCIs). BCIs allow people with motor impairments to control external devices directly with their brains using electroencephalogram (EEG) patterns that translate brain activity into control signals. Many researchers have been working to develop MI-based BCI recognition systems using various time-frequency feature extraction and classification approaches. However, the existing systems still face challenges in achieving satisfactory performance due to large amount of non-discriminative and ineffective features. To get around these problems, we suggested a multiband decomposition-based feature extraction and classification method that works well, along with a strong feature selection method for MI tasks. Our method starts by splitting the preprocessed EEG signal into four sub-bands. In each sub-band, we then used a common spatial pattern (CSP) technique to pull out narrowband-oriented useful features, which gives us a high-dimensional feature vector. Subsequently, we utilized an effective feature selection method, Relief-F, which reduces the dimensionality of the final features. Finally, incorporating advanced classification techniques, we classified the final reduced feature vector. To evaluate the proposed model, we used the three different EEG-based MI benchmark datasets, and our proposed model achieved better performance accuracy than existing systems. Our model's strong points include its ability to effectively reduce feature dimensionality and improve classification accuracy through advanced feature extraction and selection methods.

NOx concentration prediction based on multi-channel fused spectral temporal graph neural network in coal-fired power plants

Establishing an accurate and stable NOx concentration prediction model is the foundation for achieving environmental protection in coal-fired power plant denitrification. In this research, a data-driven hybrid prediction model (BMFE-MFST-GNN) is proposed to improve the prediction accuracy of inlet NOx concentrations. First, we develop the adaptive variational modal decomposition method (FEVMD) to decompose historical NOx concentration data into simple and smooth subsequences and extract time-frequency features. Second, we propose the boosting mutual information feature selection algorithm (BMIFS) to determine the best set of auxiliary variables. The delay time is calculated based on the maximal information coefficient (MIC) to reconstruct the datasets. Then, the multi-channel fused spectral temporal graph neural network (MFST-GNN) is created to build the graph feature information of decomposed subsequences and reconstructed auxiliary variables to predict the concentration subsequences. Finally, we integrate the subsequence prediction results to obtain the future NOx concentrations. The experimental results demonstrate that the proposed method outperforms several comparative models in predicting NOx concentrations.

Analysis of Vibration Characteristics of Bridge Structures under Seismic Excitation.

Bridges may undergo structural vibration responses when exposed to seismic waves. An analysis of structural vibration characteristics is essential for evaluating the safety and stability of a bridge. In this paper, a signal time-frequency feature extraction method (NTFT-ESVD) integrating standard time-frequency transformation, singular value decomposition, and information entropy is proposed to analyze the vibration characteristics of structures under seismic excitation. First, the experiment simulates the response signal of the structure when exposed to seismic waves. The results of the time-frequency analysis indicate a maximum relative error of only 1% in frequency detection, and the maximum relative errors in amplitude and time parameters are 5.9% and 6%, respectively. These simulation results demonstrate the reliability of the NTFT-ESVD method in extracting the time-frequency characteristics of the signal and its suitability for analyzing the seismic response of the structure. Then, a real seismic wave event of the Su-Tong Yangtze River Bridge during the Hengchun earthquake in Taiwan (2006) is analyzed. The results show that the seismic waves only have a short-term impact on the bridge, with the maximum amplitude of the vibration response no greater than 1 cm, and the maximum vibration frequency no greater than 0.2 Hz in the three-dimensional direction, indicating that the earthquake in Hengchun will not have any serious impact on the stability and security of the Su-Tong Yangtze River Bridge. Additionally, the reliability of determining the arrival time of seismic waves by extracting the time-frequency information from structural vibration response signals is validated by comparing it with results from seismic stations (SSE/WHN/QZN) at similar epicenter distances published by the USGS. The results of the case study show that the combination of dynamic GNSS monitoring technology and time-frequency analysis can be used to analyze the impact of seismic waves on the bridge, which is of great help to the manager in assessing structural seismic damage.

EEG Emotion Recognition Network Based on Attention and Spatiotemporal Convolution.

Human emotions are complex psychological and physiological responses to external stimuli. Correctly identifying and providing feedback on emotions is an important goal in human-computer interaction research. Compared to facial expressions, speech, or other physiological signals, using electroencephalogram (EEG) signals for the task of emotion recognition has advantages in terms of authenticity, objectivity, and high reliability; thus, it is attracting increasing attention from researchers. However, the current methods have significant room for improvement in terms of the combination of information exchange between different brain regions and time-frequency feature extraction. Therefore, this paper proposes an EEG emotion recognition network, namely, self-organized graph pesudo-3D convolution (SOGPCN), based on attention and spatiotemporal convolution. Unlike previous methods that directly construct graph structures for brain channels, the proposed SOGPCN method considers that the spatial relationships between electrodes in each frequency band differ. First, a self-organizing map is constructed for each channel in each frequency band to obtain the 10 most relevant channels to the current channel, and graph convolution is employed to capture the spatial relationships between all channels in the self-organizing map constructed for each channel in each frequency band. Then, pseudo-three-dimensional convolution combined with partial dot product attention is implemented to extract the temporal features of the EEG sequence. Finally, LSTM is employed to learn the contextual information between adjacent time-series data. Subject-dependent and subject-independent experiments are conducted on the SEED dataset to evaluate the performance of the proposed SOGPCN method, which achieves recognition accuracies of 95.26% and 94.22%, respectively, indicating that the proposed method outperforms several baseline methods.

Tunnel indirect damage identification algorithm based on service train dynamic response time series

In this study, a tunnel indirect damage identification algorithm based on a service train dynamic response time series was proposed. First, the principles and processes of the proposed algorithm are introduced. The instantaneous frequency and spectral entropy of the trend and seasonal terms of the train dynamic response decomposed by Seasonal and Trend decomposition using Loess are calculated, respectively, to further extract the time-frequency features of the signal. The calculated sequence is input into the Bi-directional Long Short-Term Memory network, and the type of damage case is the output. Second, the feasibility of the algorithm is verified using a model test. Finally, the key links in the algorithm are discussed. The results indicate that the proposed algorithm can identify damage cases with a small number of training and test sets, and the accuracy of the test set can reach 83.33%. Through time-frequency feature extraction and normalization processing, the length of the network input sequence is reduced, the training speed is increased, and damage identification can be realized without a deep network.

Enhancing the acoustic emission technique using fuzzy artificial bee colony-based deep learning for characterizing selective laser melted AlSi10Mg specimens

This article presents a classification of Acoustic Emission (AE) signals from AlSi10Mg specimens produced via Selective Laser Melting (SLM). Tensile tests characterized the mechanical properties of specimens printed in different orientations (X, Y, Z, 45°). Initially, a study quantified damage modes based on the stress-strain curve and cumulative AE energy. AE signals for each specimen (X, Y, 45°, Z), across deformation stages (elastic and plastic), and damage modes were analyzed using continuous wavelet transform to extract time-frequency features. A novel convolutional neural network, based on artificial bee colonies and fuzzy C-means, was developed for scalogram classification. Data augmentation with Gaussian white noise enhanced the approach. Cross-validation ensured robustness against overfitting and suboptimal local maxima. Evaluation metrics, including the confusion matrix, precision-recall curve, and F1 score, demonstrated the algorithm's high accuracy of 92.6%, precision-recall curve of 92.5%, and F1 score of 92.5% for AE signals based on printing direction (X, Y, 45°, Z). The study highlighted the potential for improving AE signal classification related to elastic and plastic deformation stages with 100% accuracy. For damage modes, the algorithm achieved a confusion matrix accuracy of 90.6%, a precision-recall curve of 90.4%, and an F1 score of 90.5%. This approach demonstrates high accuracy in classifying AE signals across different printing orientations, deformation stages, and damage modes of AlSi10Mg specimens manufactured through SLM.

Communication signal detection based on high-order cumulants time-frequency analysis: on the application of deep learning YOLOV5 network

The detection of communication signals in heterogeneous electromagnetic environments currently relies primarily on a one-dimensional statistical feature threshold method. However, this approach is highly sensitive to dynamic changes in the environment, fluctuations in signal-to-noise ratios, and complex noise. To address these limitations, this paper proposes a novel time-frequency diagram based on high-order accumulation for signal detection. Traditional time-frequency diagrams suffer from poor noise suppression ability and unclear features. However, higher-order cumulants can effectively overcome these shortcomings. Currently, methods based on higher-order cumulants are typically limited to one-dimensional signals. Yet, two-dimensional time-frequency signal diagrams can represent a broader array of features. This paper employs higher-order accumulation to extract time-frequency features from the received signal, thereby transforming the conventional radio detection problem into an image recognition challenge. By merging the advantages of higher-order accumulations and time-frequency diagrams, we propose the use of higher-order accumulation time-frequency diagrams for signal detection. Extensive experimental simulations demonstrate that the proposed time-frequency diagram exhibits strong anti-noise performance and effectively suppresses frequency bias from multiple perspectives. The performance of the Higher-Order Cumulant-Time Frequency (HOC-TF) indicated lower Root Mean Square Error (RMSE) compared with the Short-Time Fourier Transform-Time Frequency (STFT-TF) and Wavelet Transform-Time Frequency (WT-TF). Additionally, compared to the STFT-TF and WT-TF methodologies, the novel time-frequency diagram introduced demonstrates superior stability using the Singular Value Decomposition (SVD) method. Moreover, by combining the new time-frequency diagram with the deep learning YOLOV5 network, signal detection and modulation identification of communication signals can be achieved.

Automated identification of atrial fibrillation from single-lead ECGs using multi-branching ResNet.

Introduction: Atrial fibrillation (AF) is the most common cardiac arrhythmia, which is clinically identified with irregular and rapid heartbeat rhythm. AF puts a patient at risk of forming blood clots, which can eventually lead to heart failure, stroke, or even sudden death. Electrocardiography (ECG), which involves acquiring bioelectrical signals from the body surface to reflect heart activity, is a standard procedure for detecting AF. However, the occurrence of AF is often intermittent, costing a significant amount of time and effort from medical doctors to identify AF episodes. Moreover, human error is inevitable, as even experienced medical professionals can overlook or misinterpret subtle signs of AF. As such, it is of critical importance to develop an advanced analytical model that can automatically interpret ECG signals and provide decision support for AF diagnostics. Methods: In this paper, we propose an innovative deep-learning method for automated AF identification using single-lead ECGs. We first extract time-frequency features from ECG signals using continuous wavelet transform (CWT). Second, the convolutional neural networks enhanced with residual learning (ReNet) are employed as the functional approximator to interpret the time-frequency features extracted by CWT. Third, we propose to incorporate a multi-branching structure into the ResNet to address the issue of class imbalance, where normal ECGs significantly outnumber instances of AF in ECG datasets. Results and Discussion: We evaluate the proposed Multi-branching Resnet with CWT (CWT-MB-Resnet) with two ECG datasets, i.e., PhysioNet/CinC challenge 2017 and ECGs obtained from the University of Oklahoma Health Sciences Center (OUHSC). The proposed CWT-MB-Resnet demonstrates robust prediction performance, achieving an F1 score of 0.8865 for the PhysioNet dataset and 0.7369 for the OUHSC dataset. The experimental results signify the model's superior capability in balancing precision and recall, which is a desired attribute for ensuring reliable medical diagnoses.

Hybrid Multimodal Feature Fusion with Multi-Sensor for Bearing Fault Diagnosis.

Aiming at the traditional single sensor vibration signal cannot fully express the bearing running state, and in the high noise background, the traditional algorithm is insufficient for fault feature extraction. This paper proposes a fault diagnosis algorithm based on multi-sensor and hybrid multimodal feature fusion to achieve high-precision fault diagnosis by leveraging the operating state information of bearings in a high-noise environment to the fullest extent possible. First, the horizontal and vertical vibration signals from two sensors are fused using principal component analysis, aiming to provide a more comprehensive description of the bearing's operating condition, followed by data set segmentation. Following fusion, time-frequency feature maps are generated using a continuous wavelet transform for global time-frequency feature extraction. A first diagnostic model is then developed utilizing a residual neural network. Meanwhile, the feature data is normalized, and 28 time-frequency feature indexes are extracted. Subsequently, a second diagnostic model is constructed using a support vector machine. Lastly, the two diagnosis models are integrated to derive the final model through an ensemble learning algorithm fused at the decision level and complemented by a genetic algorithm solution to improve the diagnosis accuracy. Experimental results demonstrate the effectiveness of the proposed algorithm in achieving superior diagnostic performance with a 97.54% accuracy rate.

Reduced Biquaternion Windowed Linear Canonical Transform: Properties and Applications

The quaternion windowed linear canonical transform is a tool for processing multidimensional data and enhancing the quality and efficiency of signal and image processing; however, it has disadvantages due to the noncommutativity of quaternion multiplication. In contrast, reduced biquaternions, as a special case of four-dimensional algebra, possess unique advantages in computation because they satisfy the multiplicative exchange rule. This paper proposes the reduced biquaternion windowed linear canonical transform (RBWLCT) by combining the reduced biquaternion signal and the windowed linear canonical transform that has computational efficiency thanks to the commutative property. Firstly, we introduce the concept of a RBWLCT, which can extract the time local features of an image and has the advantages of both time-frequency analysis and feature extraction; moreover, we also provide some fundamental properties. Secondly, we propose convolution and correlation operations for RBWLCT along with their corresponding generalized convolution, correlation, and product theorems. Thirdly, we present a fast algorithm for RBWLCT and analyze its computational complexity based on two dimensional Fourier transform (2D FTs). Finally, simulations and examples are provided to demonstrate that the proposed transform effectively captures the local RBWLCT-frequency components with enhanced degrees of freedom and exhibits significant concentrations.

Multi-modal physiological time-frequency feature extraction network for accurate sleep stage classification

Sleep stage classification is essential for clinical disease diagnosis and sleep quality assessment. Most of the existing methods for sleep stage classification are based on single-channel or single-modal signal, and extract features using a single-branch, deep convolutional network, which not only hinders the capture of the diversity features related to sleep and increase the computational cost, but also has a certain impact on the accuracy of sleep stage classification. To solve this problem, this paper proposes an end-to-end multi-modal physiological time-frequency feature extraction network (MTFF-Net) for accurate sleep stage classification. First, multi-modal physiological signal containing electroencephalogram (EEG), electrocardiogram (ECG), electrooculogram (EOG) and electromyogram (EMG) are converted into two-dimensional time-frequency images containing time-frequency features by using short time Fourier transform (STFT). Then, the time-frequency feature extraction network combining multi-scale EEG compact convolution network (Ms-EEGNet) and bidirectional gated recurrent units (Bi-GRU) network is used to obtain multi-scale spectral features related to sleep feature waveforms and time series features related to sleep stage transition. According to the American Academy of Sleep Medicine (AASM) EEG sleep stage classification criterion, the model achieved 84.3% accuracy in the five-classification task on the third subgroup of the Institute of Systems and Robotics of the University of Coimbra Sleep Dataset (ISRUC-S3), with 83.1% macro F1 score value and 79.8% Cohen's Kappa coefficient. The experimental results show that the proposed model achieves higher classification accuracy and promotes the application of deep learning algorithms in assisting clinical decision-making.