Multimodal Feature Fusion Research Articles

A multi-scale and multi-modal convolutional neural network for condition monitoring of transmission line

Abstract Monitoring the fatigue damage of transmission lines is crucial for stable power system operation. However, existing model-driven methods face challenges such as high computational complexity and reliance on expert knowledge, while data-driven methods require large amounts of abnormal state data. To address these issues, a multi-scale and multi-modal convolutional neural network (CNN) is proposed for real-time condition monitoring of transmission lines. Key steps include: firstly, empirical Fourier decomposition is used to decompose the original signals, extracting multi-scale state information at different frequency scales. Then, time-domain, frequency-domain, and time–frequency domain analyses are performed on the decomposed signals to capture multi-modal information. Based on this, a multi-modal fusion network is proposed based on a CNN to extract shallow and deep features, with a fully connected layer used for multi-modal feature fusion. Notably, the algorithm is implemented on a microprocessor for practical application. Experimental results show that the proposed model achieves a diagnostic accuracy of 93.06%, outperforming classical networks. It also surpasses models trained solely on time, frequency, or time–frequency features by 25.18%, 21.8%, and 19.3%, respectively.

Enhancing anemia detection through multimodal data fusion: a non-invasive approach using EHRs and conjunctiva images

Anemia detection using multimodal approaches leverages the integration of multiple data sources, such as imaging, clinical records, and hematological parameters, to improve diagnostic accuracy. Such methods can capture the complex interplay of factors contributing to anemia, providing a more comprehensive assessment than traditional single-modality techniques. In this research, a novel deep learning multi-modal feature fusion approach is proposed for the automated detection of anemia using EHRs (Electronic Health Records), and Conjunctiva image dataset. First, EHR records are preporcessed by selecting the most appropriate features using Random Forest. The features from the conjunctiva images are extracted using RCBAM (Reverse Convolution Block Attention Mechanism). After that, GRAD-Cam algorithm is applied to calculate the pixel percentages of all the features. The output of Random Forest and GRAD-Cam algorithms is concatenated to form a multimodal fusion. The important information from the concatenated features is selected with the help of a professional healthcare consultant. The different experiments are performed on textual and image datasets individually and after concatenating. The results show that the proposed model outperforms from state-of-the-art methods with an accuracy of 95%. Despite challenges such as class imbalance and computational demands, our findings reveal substantial clinical potential, offering a patient-friendly and accessible diagnostic solution.

A novel multimodal image feature fusion mechanism:Application to rabbit liveweight estimation in commercial farms

In the realm of commercial rabbit farming, assessment of liveweight for each rabbit is crucial for production management. Traditional manual weighing methods are labor-intensive, time-consuming, and can induce stress and disease transmission. To solve this problem, this paper introduced a liveweight estimation method for rabbit utilizing near-infrared and depth image, including a segmentation stage and a weight estimation stage. Specifically, a dual-stream feature fusion mechanism was proposed to effectively integrate information from both near-infrared and depth imaging. The network was trained and validated on a dataset of 1,957 overhead images from 300 rabbits collected from a commercial farm, achieving superior performance with an R-Square (R2) of 0.95, Root Mean Square Error (RMSE) of 194.95 g, Mean Absolute Deviation (MAD) of 155.10 g, Mean Absolute Percentage Error (MAPE) of 4.08%, and Mean Coefficient of Variation (MCV) of 3.03%. Ablation experiments revealed that employing the proposed dual-stream feature fusion mechanism led to reductions of 52.87g, 38.19g, 1.01%, and 0.75% in RMSE, MAD, MAPE, and MCV, respectively, compared to not employing the mechanism. Compared to machine learning method, our proposed approach demonstrates superior performance in RMSE, MAD, MAPE, and MCV, achieving more accurate liveweight estimation for large-scale commercial farms.

Multi-modal feature fusion model based on TimesNet and T2T-ViT for ultra-short-term solar irradiance prediction

Multi-modal feature fusion model based on TimesNet and T2T-ViT for ultra-short-term solar irradiance prediction

An Adaptive Multimodal Fusion 3D Object Detection Algorithm for Unmanned Systems in Adverse Weather

Unmanned systems encounter challenging weather conditions during obstacle removal tasks. Researching stable, real-time, and accurate environmental perception methods under such conditions is crucial. Cameras and LiDAR sensors provide different and complementary data. However, the integration of disparate data presents challenges such as feature mismatches and the fusion of sparse and dense information, which can degrade algorithmic performance. Adverse weather conditions, like rain and snow, introduce noise that further reduces perception accuracy. To address these issues, we propose a novel weather-adaptive bird’s-eye view multi-level co-attention fusion 3D object detection algorithm (BEV-MCAF). This algorithm employs an improved feature extraction network to obtain more effective features. A multimodal feature fusion module has been constructed with BEV image feature generation and a co-attention mechanism for better fusion effects. A multi-scale multimodal joint domain adversarial network (M2-DANet) is proposed to enhance adaptability to adverse weather conditions. The efficacy of BEV-MCAF has been validated on both the nuScenes and Ithaca365 datasets, confirming its robustness and good generalization capability in a variety of bad weather conditions. The findings indicate that our proposed algorithm performs better than the benchmark, showing improved adaptability to harsh weather conditions and enhancing the robustness of UVs, ensuring reliable perception under challenging conditions.

A STUDENT SENTIMENT ANALYSIS METHOD BASED ON MULTIMODAL DEEP LEARNING

Based on electroencephalography (EEG) and video data, we propose a multimodal affective analysis approach in this study to examine the affective states of university students. This method is based on the findings of this investigation. The EEG signals and video data were obtained from 50 college students experiencing various emotional states, and then they were processed in great detail. The EEG signals are pre-processed to extract their multi-view characteristics. Additionally, the video data were processed by frame extraction, face detection, and convolutional neural network (CNN) operations to extract features. We take a feature splicing strategy to merge EEG and video data to produce a time series input to realize the fusion of multimodal features. This allows us to realize the fusion of multimodal features. In addition, we developed and trained a model for the classification of emotional states based on a long short-term memory network (LSTM). With the help of cross-validation, the experiments were carried out by dividing the dataset into a training set and a test set. The model’s performance was evaluated with the help of four metrics: accuracy, precision, recall, and F1-score. When compared to the single-modal method of sentiment analysis, the results demonstrate that the multimodal approach, which combines EEG and video, demonstrates considerable advantages in terms of sentiment detection. Specifically, the accuracy obtained from the multimodal approach is significantly higher. As part of its investigation, the study also investigates the respective contributions of EEG and video aspects to emotion detection. It discovers that these features complement each other in a variety of emotional states and have the potential to improve the overall recognition results. The multimodal sentiment analysis method that is based on LSTM offers a high level of accuracy and robustness when it comes to recognizing the affective states of college students. This is especially essential for enhancing the quality of education and providing support for mental health.

Self-supervised multi-modal feature fusion for predicting early recurrence of hepatocellular carcinoma

Surgical resection stands as the primary treatment option for early-stage hepatocellular carcinoma (HCC) patients. Postoperative early recurrence (ER) is a significant factor contributing to the mortality of HCC patients. Therefore, accurately predicting the risk of ER after curative resection is crucial for clinical decision-making and improving patient prognosis. This study leverages a self-supervised multi-modal feature fusion approach, combining multi-phase MRI and clinical features, to predict ER of HCC. Specifically, we utilized attention mechanisms to suppress redundant features, enabling efficient extraction and fusion of multi-phase features. Through self-supervised learning (SSL), we pretrained an encoder on our dataset to extract more generalizable feature representations. Finally, we achieved effective multi-modal information fusion via attention modules. To enhance explainability, we employed Score-CAM to visualize the key regions influencing the model’s predictions. We evaluated the effectiveness of the proposed method on our dataset and found that predictions based on multi-phase feature fusion outperformed those based on single-phase features. Additionally, predictions based on multi-modal feature fusion were superior to those based on single-modal features.

Attention-enhanced multimodal feature fusion network for clothes-changing person re-identification

Clothes-Changing Person Re-Identification is a challenging problem in computer vision, primarily due to the appearance variations caused by clothing changes across different camera views. This poses significant challenges to traditional person re-identification techniques that rely on clothing features. These challenges include the inconsistency of clothing and the difficulty in learning reliable clothing-irrelevant local features. To address this issue, we propose a novel network architecture called the Attention-Enhanced Multimodal Feature Fusion Network (AE-Net). AE-Net effectively mitigates the impact of clothing changes on recognition accuracy by integrating RGB global features, grayscale image features, and clothing-irrelevant features obtained through semantic segmentation. Specifically, global features capture the overall appearance of the person; grayscale image features help eliminate the interference of color in recognition; and clothing-irrelevant features derived from semantic segmentation enforce the model to learn features independent of the person’s clothing. Additionally, we introduce a multi-scale fusion attention mechanism that further enhances the model’s ability to capture both detailed and global structures, thereby improving recognition accuracy and robustness. Extensive experimental results demonstrate that AE-Net outperforms several state-of-the-art methods on the PRCC and LTCC datasets, particularly in scenarios with significant clothing changes. On the PRCC and LTCC datasets, AE-Net achieves Top-1 accuracy rates of 60.4% and 42.9%, respectively.

A Lightweight and Efficient Multimodal Feature Fusion Network for Bearing Fault Diagnosis in Industrial Applications.

To address the issues of single-structured feature input channels, insufficient feature learning capabilities in noisy environments, and large model parameter sizes in intelligent diagnostic models for mechanical equipment, a lightweight and efficient multimodal feature fusion convolutional neural network (LEMFN) method is proposed. Compared with existing models, LEMFN captures rich fault features at multiple scales by combining time-domain and frequency-domain signals, thereby enhancing the model's robustness to noise and improving data adaptability under varying operating conditions. Additionally, the convolutional block attention module (CBAM) and random overlapping sampling technology (ROST) are introduced, and through a feature fusion strategy, the accurate diagnosis of mechanical equipment faults is achieved. Experimental results demonstrate that the proposed method not only possesses high diagnostic accuracy and rapid convergence but also exhibits strong robustness in noisy environments. Finally, a graphical user interface (GUI)-based mechanical equipment fault detection system was developed to promote the practical application of intelligent fault diagnosis in mechanical equipment.

Integrated Sensing and Communications Toward Proactive Beamforming in mmWave V2I via Multi-Modal Feature Fusion (MMFF)

Integrated Sensing and Communications Toward Proactive Beamforming in mmWave V2I via Multi-Modal Feature Fusion (MMFF)

Visually Guided Sound Source Separation With Audio-Visual Predictive Coding.

The framework of visually guided sound source separation generally consists of three parts: visual feature extraction, multimodal feature fusion, and sound signal processing. An ongoing trend in this field has been to tailor involved visual feature extractor for informative visual guidance and separately devise module for feature fusion, while utilizing U-Net by default for sound analysis. However, such a divide-and-conquer paradigm is parameter-inefficient and, meanwhile, may obtain suboptimal performance as jointly optimizing and harmonizing various model components is challengeable. By contrast, this article presents a novel approach, dubbed audio-visual predictive coding (AVPC), to tackle this task in a parameter-efficient and more effective manner. The network of AVPC features a simple ResNet-based video analysis network for deriving semantic visual features, and a predictive coding (PC)-based sound separation network that can extract audio features, fuse multimodal information, and predict sound separation masks in the same architecture. By iteratively minimizing the prediction error between features, AVPC integrates audio and visual information recursively, leading to progressively improved performance. In addition, we develop a valid self-supervised learning strategy for AVPC via copredicting two audio-visual representations of the same sound source. Extensive evaluations demonstrate that AVPC outperforms several baselines in separating musical instrument sounds, while reducing the model size significantly. Code is available at: https://github.com/zjsong/Audio-Visual-Predictive-Coding.

Drug–drug interaction extraction based on multimodal feature fusion by Transformer and BiGRU

Understanding drug–drug interactions (DDIs) plays a vital role in the fields of drug disease treatment, drug development, preventing medical error, and controlling health care-costs. Extracting potential from biomedical corpora is a major complement of existing DDIs. Most existing DDI extraction (DDIE) methods do not consider the graph and structure of drug molecules, which can improve the performance of DDIE. Considering the different advantages of bi-directional gated recurrent units (BiGRU), Transformer, and attention mechanisms in DDIE tasks, a multimodal feature fusion model combining BiGRU and Transformer (BiGGT) is here constructed for DDIE. In BiGGT, the vector embeddings of medical corpora, drug molecule topology graphs, and structure are conducted by Word2vec, Mol2vec, and GCN, respectively. BiGRU and multi-head self-attention (MHSA) are integrated into Transformer to extract the local–global contextual DDIE features, which is important for DDIE. The extensive experiment results on the DDIExtraction 2013 shared task dataset show that the BiGGT-based DDIE method outperforms state-of-the-art DDIE approaches with a precision of 78.22%. BiGGT expands the application of multimodal deep learning in the field of multimodal DDIE.

Feature fusion and context interaction for RGB-D indoor semantic segmentation

Depth information in indoor scenes is crucial to understanding the spatial and occluding relationships of objects. Depth information contains rich geometric details and is unaffected by lighting, texture, and color, significantly enhancing traditional color images. RGB images and depth images have been widely applied in various image analysis. However, effectively leveraging the complementary information from both modalities remains a challenge. To address this issue, we propose a unified and effective feature fusion and context interaction network (FCINet). It includes three modules. Firstly, we construct a multi-modal feature fusion module (MFFM), which can achieve the aggregation of two modalities along spatial and channel dimensions. Secondly, we construct a global and local information interaction context module (GLCM) to encode rich semantic information. Finally, we construct a feature alignment and fusion module (FAFM), which integrates upsampled low-level features with high-level features to mitigate spatial information loss. Experimental results indicate that the model achieved favorable outcomes on both the NYU Depth v2 dataset and the more complex SUN RGB-D dataset.

DSEM-NeRF: Multimodal feature fusion and global–local attention for enhanced 3D scene reconstruction

3D scene understanding often faces the problems of insufficient detail capture and poor adaptability to multi-view changes. To this end, we proposed a NeRF-based 3D scene understanding model DSEM-NeRF, which effectively improves the reconstruction quality of complex scenes through multimodal feature fusion and global–local attention mechanism. DSEM-NeRF extracts multimodal features such as color, depth, and semantics from multi-view 2D images, and accurately captures key areas by dynamically adjusting the importance of features. Experimental results show that DSEM-NeRF outperforms many existing models on the LLFF and DTU datasets, with PSNR reaching 20.01, 23.56, and 24.58 respectively, and SSIM reaching 0.834. In particular, it shows strong robustness in complex scenes and multi-view changes, verifying the effectiveness and reliability of the model.

Multi-Modal Object Detection Method Based on Dual-Branch Asymmetric Attention Backbone and Feature Fusion Pyramid Network

With the simultaneous acquisition of the infrared and optical remote sensing images of the same target becoming increasingly easy, using multi-modal data for high-performance object detection has become a research focus. In remote sensing multi-modal data, infrared images lack color information, it is hard to detect difficult targets with low contrast, and optical images are easily affected by illuminance. One of the most effective ways to solve this problem is to integrate multi-modal images for high-performance object detection. The challenge of fusion object detection lies in how to fully integrate multi-modal image features with significant modal differences and avoid introducing interference information while taking advantage of complementary advantages. To solve these problems, a new multi-modal fusion object detection method is proposed. In this paper, the method is improved in terms of two aspects: firstly, a new dual-branch asymmetric attention backbone network (DAAB) is designed, which uses a semantic information supplement module (SISM) and a detail information supplement module (DISM) to supplement and enhance infrared and RGB image information, respectively. Secondly, we propose a feature fusion pyramid network (FFPN), which uses a Transformer-like strategy to carry out multi-modal feature fusion and suppress features that are not conducive to fusion during the fusion process. This method is a state-of-the-art process for both FLIR-aligned and DroneVehicle datasets. Experiments show that this method has strong competitiveness and generalization performance.

Metal surface crack depth laser acoustic emission detection method based on multivariate feature adaptive extraction and cross-modal interaction fusion

Surface crack depth detection of metal structures is of great significance to ensure the safe of equipment. However, the commonly used detection methods currently suffer from sensitivity and imaging dimension limitations, which make it difficult to ensure accuracy and feature extraction. To this end, a laser acoustic emission crack depth detection method that combines modal decomposition and deep learning is proposed. The new method is more sensitive to elastic wave changes that detect the depth of microscopic cracks and shows higher accuracy than traditional detection methods. At the same time, frequency-adaptive feature mode decomposition of multivariate signals and color multimodal feature fusion are used to solve the difficulty of time–frequency domain feature mining in current research. The crack depth detection network CDDNet is constructed using an attention-based fusion backbone, a novel cross-modal interaction fusion, and a hierarchical information fusion strategy. Its transfer learning achieves a crack detection accuracy level of 0.05 mm and outperforms other models in terms of accuracy, reaching 93.6 %. This paper shows that the proposed transfer learning laser acoustic emission technology and supporting data processing methods can effectively establish a connection between the experimental and simulation datasets, solving the current challenges of crack depth detection in terms of micro-precision and data processing.

Augmenting Multimodal Content Representation with Transformers for Misinformation Detection

Information sharing on social media has become a common practice for people around the world. Since it is difficult to check user-generated content on social media, huge amounts of rumors and misinformation are being spread with authentic information. On the one hand, most of the social platforms identify rumors through manual fact-checking, which is very inefficient. On the other hand, with an emerging form of misinformation that contains inconsistent image–text pairs, it would be beneficial if we could compare the meaning of multimodal content within the same post for detecting image–text inconsistency. In this paper, we propose a novel approach to misinformation detection by multimodal feature fusion with transformers and credibility assessment with self-attention-based Bi-RNN networks. Firstly, captions are derived from images using an image captioning module to obtain their semantic descriptions. These are compared with surrounding text by fine-tuning transformers for consistency check in semantics. Then, to further aggregate sentiment features into text representation, we fine-tune a separate transformer for text sentiment classification, where the output is concatenated to augment text embeddings. Finally, Multi-Cell Bi-GRUs with self-attention are used to train the credibility assessment model for misinformation detection. From the experimental results on tweets, the best performance with an accuracy of 0.904 and an F1-score of 0.921 can be obtained when applying feature fusion of augmented embeddings with sentiment classification results. This shows the potential of the innovative way of applying transformers in our proposed approach to misinformation detection. Further investigation is needed to validate the performance on various types of multimodal discrepancies.

Joint segmentation of tumors in 3D PET-CT images with a network fusing multi-view and multi-modal information

Objective. Joint segmentation of tumors in positron emission tomography-computed tomography (PET-CT) images is crucial for precise treatment planning. However, current segmentation methods often use addition or concatenation to fuse PET and CT images, which potentially overlooks the nuanced interplay between these modalities. Additionally, these methods often neglect multi-view information that is helpful for more accurately locating and segmenting the target structure. This study aims to address these disadvantages and develop a deep learning-based algorithm for joint segmentation of tumors in PET-CT images.Approach. To address these limitations, we propose the Multi-view Information Enhancement and Multi-modal Feature Fusion Network (MIEMFF-Net) for joint tumor segmentation in three-dimensional PET-CT images. Our model incorporates a dynamic multi-modal fusion strategy to effectively exploit the metabolic and anatomical information from PET and CT images and a multi-view information enhancement strategy to effectively recover the lost information during upsamping. A Multi-scale Spatial Perception Block is proposed to effectively extract information from different views and reduce redundancy interference in the multi-view feature extraction process.Main results. The proposed MIEMFF-Net achieved a Dice score of 83.93%, a Precision of 81.49%, a Sensitivity of 87.89% and an IOU of 69.27% on the Soft Tissue Sarcomas dataset and a Dice score of 76.83%, a Precision of 86.21%, a Sensitivity of 80.73% and an IOU of 65.15% on the AutoPET dataset.Significance. Experimental results demonstrate that MIEMFF-Net outperforms existing state-of-the-art models which implies potential applications of the proposed method in clinical practice.

WCNN-RSN: a novel fault diagnosis method for rolling bearing using multimodal feature fusion

Abstract Fault diagnosis of rolling bearings is significant for the safe operation of engineering equipment. Many intelligent diagnostic methods have been successfully developed. However, it is often susceptible to noisy environments. Therefore, the paper proposes a rolling bearing fault diagnosis method based on multimodal information fusion in time and time–frequency domains by combining an improved 1D-convolutional neural network (CNN) with ResNet50 wavelet improved CNN-ResNet (WCNN-RSN). The algorithm employs the Multi-Head Attention (MHA) mechanism to complementarily fuse fault features in different scales, achieving fault diagnosis by fully extracting fault features. The experimental results show that the diagnostic effect of WCNN-RSN is better than that of the comparison methods under noise interference, which proves that the proposed method possesses good anti-noise and generalization ability.

Multimodal and multiscale feature fusion for weakly supervised video anomaly detection

Weakly supervised video anomaly detection aims to detect anomalous events with only video-level labels. In the absence of boundary information for anomaly segments, most existing methods rely on multiple instance learning. In these approaches, the predictions for unlabeled video snippets are guided by the classification of labeled untrimmed videos. However, these methods do not account for issues such as video blur and visual occlusion, which can hinder accurate anomaly detection. To address these issues, we propose a novel weakly supervised video anomaly detection method that fuses multimodal and multiscale features. Firstly, RGB and optical flow snippets are input into pre-trained I3D to extract appearance and motion features. Then, we introduce an Attention De-redundancy (AD) module, which employs an attention mechanism to filter out task-irrelevant redundancy in these appearance and motion features. Next, to mitigate the effects of video blurring and visual occlusion, we propose a Multi-scale Feature Learning module. This module captures long-term and short-term temporal dependencies among video snippets to provide global and local guidance for blurred or occluded video snippets. Finally, to effectively utilize the discriminative features of different modalities, we propose an Adaptive Feature Fusion module. This module adaptively fuses appearance and motion features based on their respective feature weights. Extensive experimental results demonstrate that our proposed method outperforms mainstream unsupervised and weakly supervised methods in terms of AUC. Specifically, our proposed method achieves 97.00% AUC and 85.31% AUC on two benchmark datasets, i.e., ShanghaiTech and UCF-Crime, respectively.