Final Classification Accuracy Research Articles

Traceability of Microplastic Fragments from Waste Plastic Express Packages Using Near-Infrared Spectroscopy Combined with Chemometrics.

The pollution from waste plastic express packages (WPEPs), especially microplastic (MP) fragments, caused by the blowout development of the express delivery industry has attracted widespread attention. On account of the variety of additives, strong complexity, and high diversity of plastic express packages (PEPs), the multi-class classification of WPEPs is a typical large-class-number classification (LCNC). The traceability and identification of microplastic fragments from WPEPs is very challenging. An effective chemometric method for large-class-number classification would be very beneficial for the comprehensive treatment of WPEP pollution through the recycling and reuse of waste plastic express packages, including microplastic fragments and plastic debris. Rather than using the traditional one-against-one (OAO) and one-against-all (OAA) dichotomies, an exhaustive and parallel half-against-half (EPHAH) decomposition, which overcomes the defects of the OAO's classifier learning limitations and the OAA's data proportion imbalance, is proposed for feature selection. EPHAH analysis, combined with partial least squares discriminant analysis (PLS-DA) for large-class-number classification, was performed on 750 microplastic fragments of polyethylene WPEPs from 10 major courier companies using near-infrared (NIR) spectroscopy. After the removal of abnormal samples through robust principal component analysis (RPCA), the root mean square error of cross-validation (RMSECV) value for the model was reduced to 0.01, which was 21.5% lower than that including the abnormal samples. The best models of PLS-DA were obtained using SNV combined with SG-17 smoothing and 2D (SNV+SG-17+2D); the latent variables (LVs), the error rates of Monte Carlo cross-validation (ERMCCVs), and the final classification accuracies were 6.35, 0.155, and 88.67% for OAO-PLSDA; 5.37, 0.103, and 87.33% for OAA-PLSDA; and 3.12, 0.054, and 96.00% for EPHAH-PLSDA. The results showed that the EPHAH strategy can completely learn the complex LCNC decision boundaries for 10 classes, effectively break the tie problem, and greatly improve the voting resolution, thereby demonstrating significant superiority to both the OAO and OAA strategies in terms of classification accuracy. Meanwhile, PLS-DA further maximized the covariance and data interpretation abilities between the potential variables and categories of microplastic debris, thereby establishing an ideal performance identification model with a recognition rate of 96.00%.

Identification of rice disease under complex background based on PSOC-DRCNet

Rice is a crucial agricultural crop, yet it frequently suffers from various diseases, leading to decreased yields and, in severe cases, crop failure. Diseases significantly affect rice growth and yield, resulting in economic losses and food security challenges. The role of image recognition in identifying rice diseases is critical in agricultural production. It enables automated and efficient detection of rice diseases, which is essential for effective management, ensuring food security and sustainable agriculture. To address issues like background noise and edge blurring in rice disease image capture, as well as challenges in determining the optimal learning rate during the training of traditional rice disease recognition networks, a novel method based on PSOC-DRCNet is proposed for rice disease recognition.. First, tto solve the problem of background interference, Dual Mode Attention (DMA) is proposed to adaptively capture meaningful regions in rice disease images. Second, the Residual Adaptive Block(RAB) is proposed, which utilizes dimensional changes and channel attention to solve edge blur problems. Then, a Cross entropy and regularized mixed Loss function (CerLoss), is proposed to optimize the learning strategy of the model in the process of processing datasets and enhance the performance and generalization ability of the model to avoid overfitting problems. Ultimately, In response to the cumbersome problem of finding the optimal learning rate, we propose using Particle Swarm Optimization Chameleon (PSOC) to find the optimal learning rate and train the PSOC-DRCNet model on our custom dataset and compare it with other existing methods and the final average classification accuracy of PSOC-DRCNet is 93.88% with an F1 score of 0.940. We compare it with other existing methods. It is proved that the average classification accuracy of our model under hyper-parameter unification is 92.65% F1 score is 0.928. We validated the PSOC-DRCNet by conducting comparative analyses with other models and through generalization experiments and module effectiveness tests. Additionally, the practicality of PSOC-DRCNet was confirmed through its application in real-world scenarios. The methods proposed in this paper successfully enable the identification of various diseases in rice leaves, offering a practical solution for incorporating deep learning into the agricultural production process. Furthermore, these findings serve as a valuable reference for disease identification in other crops.

Image Classification and Restoration of Ancient Textiles Based on Convolutional Neural Network

Ancient textile images have a variety of styles and themes, and the classification of different types of textiles provides a reliable reference for the protection and restoration of cultural relics. Due to the low efficiency of traditional classification methods and the low accuracy of classification, the image restoration of textiles takes longer and the repair effect is poor. Therefore, this paper takes ancient textile images as the research object and selects YOLOv4–ViT collaborative identification network (YOLOv4–ViT network) and generative adversarial networks (GAN) restoration model from a variety of network models to classify and restore ancient textile images. In this work, YOLOv4–ViT network is used to recognize and classify pattern elements in ancient textile images. Then, according to the classification results, restoration training of ancient textiles was carried out using an improved GAN restoration model, for which the final classification accuracy reached 92.78% and the repair result even took only 1.5 s. On this basis, a reliable retrieval and restoration system is designed to realize the repair of damaged textile images, reduce the difficulty of repair, and help users retrieve and browse different categories of ancient textile images, thus solve the problems of slow retrieval speed in traditional retrieval methods and poor restoration effect of ancient textile images.

High Precision Cervical Precancerous Lesion Classification Method Based on ConvNeXt.

Traditional cervical cancer diagnosis mainly relies on human papillomavirus (HPV) concentration testing. Considering that HPV concentrations vary from individual to individual and fluctuate over time, this method requires multiple tests, leading to high costs. Recently, some scholars have focused on the method of cervical cytology for diagnosis. However, cervical cancer cells have complex textural characteristics and small differences between different cell subtypes, which brings great challenges for high-precision screening of cervical cancer. In this paper, we propose a high-precision cervical cancer precancerous lesion screening classification method based on ConvNeXt, utilizing self-supervised data augmentation and ensemble learning strategies to achieve cervical cancer cell feature extraction and inter-class discrimination, respectively. We used the Deep Cervical Cytological Levels (DCCL) dataset, which includes 1167 cervical cytology specimens from participants aged 32 to 67, for algorithm training and validation. We tested our method on the DCCL dataset, and the final classification accuracy was 8.85% higher than that of previous advanced models, which means that our method has significant advantages compared to other advanced methods.

A design and implementation of multi-character classification scheme based on motor imagery EEG signals

In the field of brain-to-text communication, it is difficult to finish highly dexterous behaviors of writing multi-character by motor-imagery-based brain–computer interface (MI-BCI), setting a barrier to restore communication in people who have lost the ability to move and speak. In this paper, we design and implement a multi-character classification scheme based on 29 characters of motor imagery (MI) electroencephalogram (EEG) signals, which contains 26 English letters and 3 punctuation marks. Firstly, we design a novel experimental paradigm to increase the variety of BCI inputs by asking subjects to imagine the movement of writing 29 characters instead of gross motor skills such as reaching or grasping. Secondly, because of the high dimension of EEG signals, we adopt power spectral density (PSD), principal components analysis (PCA), kernel principal components analysis (KPCA) respectively to decompose EEG signals and extract feature, and then test the results with pearson product-moment correlation coefficient (PCCs). Thirdly, we respectively employ k-nearest neighbor (kNN), support vector machine (SVM), extreme learning machine (ELM) and light gradient boosting machine (LightGBM) to classify 29 characters and compare the results. We have implemented a complete scheme, including paradigm design, signal acquisition, feature extraction and classification, which can effectively classify 29 characters. The experimental results show that the KPCA has the best feature extraction effect and the kNN has the highest classification accuracy, with the final classification accuracy reaching 96.2%, which is better than other studies.

ASCNet: Attention Mechanism and Self-Calibration Convolution Fusion Network for X-ray Femoral Fracture Classification

X-ray examinations are crucial for fracture diagnosis and treatment. However, some fractures do not present obvious imaging feature in early X-rays, which can result in misdiagnosis. Therefore, an ASCNet model is proposed in this study for X-ray femoral fracture classification. This model adopts the self-calibration convolution method to obtain more discriminative feature representation. This convolutional way can enable each spatial location to adaptively encode the context information of distant regions and make the model obtain some characteristic information hidden in X-ray images. Additionaly, the ASCNet model integrates the convolutional block attention module and coordinate attention module to capture different information from space and channels to fully obtain the apparent fracture features in X-ray images. Finally, the effectiveness of the proposed model is verified using the femoral fracture dataset. The final classification accuracy and AUC value of the ASCNet are 0.9286 and 0.9720, respectively. The experimental results demonstrate that the ASCNet model performs better than ResNet50 and SCNet50. Furthermore, the proposed model presents specific advantages in recognizing occult fractures in X-ray images.

Interpretable Artificial Intelligence (AI) Differentiates Prefibrotic Primary Myelofibrosis (prePMF) from Essential Thrombocythemia (ET): A Multi-Center Study of a New Clinical Decision Support Tool

Introduction Overlapping clinical, molecular, and histopathological characteristics pose challenges in differentiating prePMF from ET. The median overall survival, however, significantly differs between prePMF and ET (11.9 vs 22.2 years, Jeryczynski, 2017). The difference in survival highlights the need to distinguish between these two myeloproliferative neoplasms (MPNs) to select disease-specific therapeutic options. This area of unmet need often requires expert assessment at high-volume academic institutions to render a definitive diagnosis. Our aim in this study is to develop and validate a biologically-motivated AI algorithm to rapidly, accurately, and inexpensively diagnose prePMF and ET directly from diagnostic bone marrow (BM) biopsy digital whole-slide images (WSI). Methods Patients with a clinical/histopathological diagnosis of prePMF or ET as determined by the International Consensus Classification of Myeloid Neoplasms were identified at the University of Florence, Italy (Florence) between 06/2007 and 05/2023 and Moffitt Cancer Center, Tampa, FL (Moffitt) between 01/2013 and 01/2022. Diagnostic H&E-stained BM biopsy slides were digitized using Aperio AT2 slide scanners (Leica Biosystems, Deer Park, IL) at each institution . The training cohort comprised of 200 (100 prePMF / 100 ET) patients from Florence, and the external test cohort entailed 26 (6 prePMF / 20 ET) patients from Moffitt. In total, the resultant model was trained on 32,226 patient-derived WSI. Our chosen pretrained neural network, RetCCL, was previously trained on 32,000 diagnostic WSIs to potentially represent a histologically-informed model (Wang, 2023). BM WSI were tessellated into representative image tiles extracted at 10x magnification (302 microns per image dimensions) for model training. Finally, a prediction upon each patient's WSI was calculated by attention-based multiple instance learning, which is a method that automatically assigns a numeric weight to an image portion representing its relative importance to the classification task. Model performance was assessed utilizing the area under the receiver operator curve (AUC). The cutoff threshold for diagnosis classification was determined by maximizing Youden's Index. For qualitative assessment, attention scores were plotted as a heatmap across the BM WSI and reviewed for morphological features by an expert hematopathologist. Custom scripts were written using our open-source AI framework, Slideflow (Dolezal, 2021). Model development was performed on the Minerva High Performance Computer at Mount Sinai Hospital. Evaluation time upon a single WSI was estimated using a consumer-grade computer with an NVIDIA RTX 3080 graphics processing unit. Results Within the training cohort, 5-fold cross validation resulted in a mean AUC of 0.90 and standard deviation of 0.04. A final locked model re-trained on the entire training cohort resulted in an AUC of 0.90 upon evaluation of the test cohort ( Figure 1). We optimized the classification threshold to balance sensitivity and specificity; the final diagnostic classification accuracy on the test cohort was 92.3% with a sensitivity and specificity for prePMF diagnosis of 66.6% and 100%, respectively. Upon review of the slides with highest prediction value per class, attention heatmaps highlighted the model's reliance on areas of cellular marrow without reliance on image artifacts or background ( Figure 2). Using affordable consumer-grade hardware, evaluation upon a previously unseen WSI was completed in approximately 6.1 seconds (4.9 for preprocessing and 1.2 for evaluation). Conclusion We developed a novel AI model with high accuracy for distinguishing between prePMF and ET in distinct clinical cohorts. To our knowledge, this study represents the largest image-based AI study within MPNs with external validation. Our proposed model may assist clinicians in appropriately identifying patient cohorts who would benefit from disease-specific therapies or enrollment onto clinical trials. We imagine that a potential high-speed, low-cost algorithm may reliably distinguish prePMF from ET patients with high specificity which can be democratized to the MPN clinical community in routine practice and drive clinical trial accrual for biologically rational novel therapeutics.

MMTFN: Multi‐modal multi‐scale transformer fusion network for Alzheimer's disease diagnosis

AbstractAlzheimer's disease (AD) is a severe neurodegenerative disease that can cause dementia symptoms. Currently, most research methods for diagnosing AD rely on fusing neuroimaging data of different modalities to exploit their heterogeneity and complementarity. However, effectively using such multi‐modal information to construct fusion methods remains a challenging problem. To address this issue, we propose a multi‐modal multi‐scale transformer fusion network (MMTFN) for computer‐aided diagnosis of AD. Our network comprises 3D multi‐scale residual block (3DMRB) layers and the Transformer network that jointly learns potential representations of multi‐modal data. The 3DMRB with multi‐scale aggregation efficiently extracts local abnormal information related to AD in the brain. We conducted five experiments to validate our model using MRI and PET images of 720 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI). The experimental results show that our proposed network outperformed existing models, achieving a final classification accuracy of 94.61% for AD and Normal Control.

Automated Age-Related Macular Degeneration Detector on Optical Coherence Tomography Images Using Slice-Sum Local Binary Patterns and Support Vector Machine.

Artificial intelligence has revolutionised smart medicine, resulting in enhanced medical care. This study presents an automated detector chip for age-related macular degeneration (AMD) using a support vector machine (SVM) and three-dimensional (3D) optical coherence tomography (OCT) volume. The aim is to assist ophthalmologists by reducing the time-consuming AMD medical examination. Using the property of 3D OCT volume, a modified feature vector connected method called slice-sum is proposed, reducing computational complexity while maintaining high detection accuracy. Compared to previous methods, this method significantly reduces computational complexity by at least a hundredfold. Image adjustment and noise removal steps are excluded for classification accuracy, and the feature extraction algorithm of local binary patterns is determined based on hardware consumption considerations. Through optimisation of the feature vector connection method after feature extraction, the computational complexity of SVM detection is significantly reduced, making it applicable to similar 3D datasets. Additionally, the design supports model replacement, allowing users to train and update classification models as needed. Using TSMC 40 nm CMOS technology, the proposed detector achieves a core area of 0.12 mm2 while demonstrating a classification throughput of 8.87 decisions/s at a maximum operating frequency of 454.54 MHz. The detector achieves a final testing classification accuracy of 92.31%.

Automatic labelling for semantic segmentation of VHR satellite images: Application of airborne laser scanner data and object-based image analysis

The application of deep learning methods to remote sensing data has produced good results in recent studies. A promising application area is automatic land cover classification (semantic segmentation) from very high-resolution satellite imagery. However, the deep learning methods require large, labelled training datasets that are suitable for the study area. Map data can be used as training data, but it is often insufficiently detailed for very high-resolution satellite imagery. National airborne laser scanner (lidar) datasets provide additional details and are available in many countries. Successful land cover classifications from lidar datasets have been reached, e.g., by object-based image analysis. In the present study, we investigated the feasibility of using airborne laser scanner data and object-based image analysis to automatically generate labelled training data for a deep neural network -based land cover classification of a VHR satellite image. Input data for the object-based classification included digital surface models, intensity and pulse information derived from the lidar data. The resulting land cover classification was then utilized as training data for deep learning. A state-of-the-art deep learning architecture, UnetFormer, was trained and applied to the land cover classification of a WorldView-3 stereo dataset. For the semantic segmentation, three different input data composites were produced using the red, green, blue, NIR and digital surface model bands derived from the satellite data. The quality of the generated training data and the semantic segmentation results was estimated using an independent test set of ground truth points. The results show that final satellite image classification accuracy (94–96%) close to the training data accuracy (97%) was obtained. It was also demonstrated that the resulting maps could be used for land cover change detection.

Identification and classification of metal copper based on laser-induced breakdown spectroscopy

Precious and half-precious metals are widely used in various fields, which makes it of great significance to recycle them, and copper was taken as an example for the investigation in this paper. A system based on laser-induced breakdown spectroscopy combined with machine learning algorithms was developed and employed in the lab to identify and classify several metal devices that contain copper element. According to the obtained emission spectra, 36 characteristic spectral lines of copper element are observed in the spectrogram of high-purity copper, as well as some metallic elements including Zn, Ca, Mg, and Na that also appeared. Moreover, eight types of similar metal devices containing copper element which are common in life (electrode, copper plug, copper tape, carbon brush, wire, circuit board, gasket, and coil) were selected to perform spectral analysis. Rough classification can be achieved by observing the spectra of eight metal devices. The effective classification process of metal devices was implemented by conducting principal component analysis, which built a model to reduce the dimension of spectral data for classification. Several samples are distributed at different positions in the principal component space, which is established based on the three principal components as the coordinate axis. K-nearest neighbors were employed to verify the classification effectiveness, acquiring the final classification accuracy of 99%. The results show that the development system has a broad development prospect for identifying metal copper and classifying metal devices that contain copper element.

Leaking Gas Source Tracking for Multiple Chemical Parks within An Urban City

Sudden air pollution accidents (explosions, fires, leaks, etc.) in chemical industry parks may result in great harm to people’s lives, property, and the ecological environment. A gas tracking network can monitor hazardous gas diffusion using traceability technology combined with sensors distributed within the scope of a chemical industry park. Such systems can automatically locate the source of pollutants in a timely manner and notify relevant departments to take major hazards into their control. However, tracing the source of the leak in a large area is still a tough problem, especially within an urban area. In this paper, the diffusion of 79 potential leaking sources with consideration of different weather conditions and complex urban terrain is simulated by AERMOD. Only 61 sensors are used to monitor the gas concentration within such a large scale. A fully connected network trained with a hybrid strategy is proposed to trace the leaking source effectively and robustly. Our proposed model reaches a final classification accuracy of 99.14%.

Quantitative Classification Model of Composite Product Image Based on Event-Related Potential

As an important research tool in neuroscience, event-related potential (ERP) technology enables in-depth analysis of the consumer’s product image cognition process and complements and verifies the product image cognition model at the ERP level. It provides an important theoretical basis for systematically capturing product image and improvement of the product image cognitive model. In this work, the correlation between ERP data, product image word pairs and the degree of semantic match with the product is investigated, and a support vector machine algorithm is selected to build a classification model with physiological data (behavioral data + ERP data) as the independent variable and the degree of semantic match with the product image as the dependent variable. By adjusting the model parameters, the final classification accuracy reaches 95.667%, which shows that the model has some reliability and is a viable research method for ERP-based product image researchers in the future.

Research on Intelligent Monitoring of Boring Bar Vibration State Based on Shuffle-BiLSTM.

Due to its low stiffness, the boring bar used in deep-hole-boring is prone to violent vibration during the cutting process. It is often inaccurate and inefficient to judge the vibration state of the boring bar through artificial experience. To detect the change of the vibration state of the boring bar over time, guide the adjustment of the processing parameters, and avoid wastage of the workpiece and the loss of equipment, it is particularly important to intelligently monitor the vibration state of the boring bar during processing. In this paper, the boring bar is taken as the research object, and an intelligent monitoring technology of the boring bar's vibration state based on deep learning is proposed. Based on grouping convolution, channel shuffle, and BiLSTM, a shuffle-BiLSTM NET model is constructed, which is both lightweight and has a high classification accuracy. The boring experiment platform is built, and 192 groups of cutting experiments are carried out. The three-way acceleration and sound pressure signals are collected, and the signals are processed by smoothed pseudo-Wigner-Ville distribution. The original signals are transformed into a 256 × 256 × 3 matrix obtained by a two-dimensional time-frequency spectrum diagram. The matrix is input into the model to recognize the boring bar's vibration state. The final classification accuracy is 91.2%. A variety of typical deep learning models are introduced for performance comparison, which proves the superiority of the models and methods used in this paper.

Tree Species Classification in UAV Remote Sensing Images Based on Super-Resolution Reconstruction and Deep Learning

We studied the use of self-attention mechanism networks (SAN) and convolutional neural networks (CNNs) for forest tree species classification using unmanned aerial vehicle (UAV) remote sensing imagery in Dongtai Forest Farm, Jiangsu Province, China. We trained and validated representative CNN models, such as ResNet and ConvNeXt, as well as the SAN model, which incorporates Transformer models such as Swin Transformer and Vision Transformer (ViT). Our goal was to compare and evaluate the performance and accuracy of these networks when used in parallel. Due to various factors, such as noise, motion blur, and atmospheric scattering, the quality of low-altitude aerial images may be compromised, resulting in indistinct tree crown edges and deficient texture. To address these issues, we adopted Real-ESRGAN technology for image super-resolution reconstruction. Our results showed that the image dataset after reconstruction improved classification accuracy for both the CNN and Transformer models. The final classification accuracies, validated by ResNet, ConvNeXt, ViT, and Swin Transformer, were 96.71%, 98.70%, 97.88%, and 98.59%, respectively, with corresponding improvements of 1.39%, 1.53%, 0.47%, and 1.18%. Our study highlights the potential benefits of Transformer and CNN for forest tree species classification and the importance of addressing the image quality degradation issues in low-altitude aerial images.

A ReRAM-Based Convolutional Neural Network Accelerator Using the Analog Layer Normalization Technique

This paper presents a ReRAM-based convolutional neural network (CNN) accelerator with a new analog layer normalization (ALN) technique. The proposed ALN can be used to effectively reduce the effect of the conductance variation is ReRAM devices by normalizing the outputs of the vector-matrix multiplication (VMM) in the charge domain. The ALN achieves high energy and hardware efficiencies because it directly processes normalization of the VMM outputs without storing their values in memory and is merged into the neuron circuit of the accelerator. To verify the effect of the ALN through experiments, a VMM accelerator that consists of two 25 × 25 sized ReRAM arrays and peripheral circuits with ALN is used for a convolution layer with digital signal processing (DSP) in a field programmable gate array (FPGA). The MNIST dataset is used to train and inference a CNN employing two VMM accelerators that work as convolution layers in a pipelined manner. Despite the conductance variation of the ReRAM devices, the ALN successfully stabilizes the output distribution of the convolution layer, which improves the classification accuracy of the network. A final classification accuracy for the MNIST and Fashion-MNIST datasets of 96.2% and 83.1% is achieved, respectively, with an energy efficiency of 9.94 tera-operations per second per Watt (TOPS/W).

A Context-focused Attention Evolution Model for Aspect-based Sentiment Classification

Due to their inherent capability in the semantic alignment of aspects and their context words, Attention and Long-Short-Term-Memory (LSTM) mechanisms are widely adopted for Aspect-Based Sentiment Classification (ABSC) tasks. Instead, it is challenging to handle long-range word dependencies on multiple entities due to the deficiency in attention mechanisms. To solve this problem, we propose a Context-Focused Aspect-Based Network to align attention before LSTM, making the model focus more on aspect-related words and ignore irrelevant words, improving the accuracy of final classification. This can either alleviate attention distraction or reinforce the text representation ability. Experiments on two benchmark datasets show that the results achieve respectable performance compared to the state-of-the-art methods available in ABSC. Our approach has the potential to improve classification accuracy by adaptively adjusting the focus on context.

Multi-stage biomedical feature selection extraction algorithm for cancer detection

Cancer is a significant cause of death worldwide. Early cancer detection is greatly aided by machine learning and artificial intelligence (AI) to gene microarray data sets (microarray data). Despite this, there is a significant discrepancy between the number of gene features in the microarray data set and the number of samples. Because of this, it is crucial to identify markers for gene array data. Existing feature selection algorithms, however, generally use long-standing, are limited to single-condition feature selection and rarely take feature extraction into account. This work proposes a Multi-stage algorithm for Biomedical Deep Feature Selection (MBDFS) to address this issue. In the first, three feature selection techniques are combined for thorough feature selection, and feature subsets are obtained; in the second, an unsupervised neural network is used to create the best representation of the feature subset to enhance final classification accuracy. Using a variety of metrics, including a comparison of classification results before and after feature selection and the performance of alternative feature selection methods, we evaluate MBDFS's efficacy. The experiments demonstrate that although MBDFS uses fewer features, classification accuracy is either unchanged or enhanced.

A CNN-Based E-Nose Using Time Series Features for Food Freshness Classification

In this article, a convolutional neural network (CNN)-based electronic nose system is proposed, which utilizes time series features to classify the freshness of food. The electronic nose is composed of three elements: a highly sensitive miniaturized micro-electromechanical systems (MEMS) metal–oxide (MOX)–semiconductor gas sensor array, a complementary metal–oxide–semiconductor (CMOS) integrated circuit, and a CNN-based classification unit. The gas sensors exhibit exceptional sensitivity, low power consumption, and a small size. The interface circuit for sensor readout is fabricated using a standard 180-nm CMOS process, with an area of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1.5\times1.5$ </tex-math></inline-formula> mm. The interface chip mainly consists of an analog-to-digital converter (ADC), an electrically erasable programmable read-only memory (EEPROM), an oscillator (OSC), power management, a heater driver, and a digital control circuit. Fixed <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R}{(}{t}{)}$ </tex-math></inline-formula> exposures are taken at specific intervals under gas input and output conditions. Time series features are extracted from a diverse set of sensor signals, which allows the subtle differences in various food odors at different freshness levels to be identified. The incorporation of time series feature extraction extends the data features, resulting in enhanced classification accuracy with a limited number of sensors. An abstract odor map input to the CNN is formed by combining the steady-state and transient information of the gas sensor response. The final freshness classification accuracy of 20 types of foods is 97.3%. The accuracy is improved by 6.5% after the implementation of time series feature extraction, demonstrating the significance of instantaneous fluctuation information for classification accuracy.

A Fusion-Assisted Multi-Stream Deep Learning and ESO-Controlled Newton–Raphson-Based Feature Selection Approach for Human Gait Recognition

The performance of human gait recognition (HGR) is affected by the partial obstruction of the human body caused by the limited field of view in video surveillance. The traditional method required the bounding box to recognize human gait in the video sequences accurately; however, it is a challenging and time-consuming approach. Due to important applications, such as biometrics and video surveillance, HGR has improved performance over the last half-decade. Based on the literature, the challenging covariant factors that degrade gait recognition performance include walking while wearing a coat or carrying a bag. This paper proposed a new two-stream deep learning framework for human gait recognition. The first step proposed a contrast enhancement technique based on the local and global filters information fusion. The high-boost operation is finally applied to highlight the human region in a video frame. Data augmentation is performed in the second step to increase the dimension of the preprocessed dataset (CASIA-B). In the third step, two pre-trained deep learning models-MobilenetV2 and ShuffleNet-are fine-tuned and trained on the augmented dataset using deep transfer learning. Features are extracted from the global average pooling layer instead of the fully connected layer. In the fourth step, extracted features of both streams are fused using a serial-based approach and further refined in the fifth step by using an improved equilibrium state optimization-controlled Newton-Raphson (ESOcNR) selection method. The selected features are finally classified using machine learning algorithms for the final classification accuracy. The experimental process was conducted on 8 angles of the CASIA-B dataset and obtained an accuracy of 97.3, 98.6, 97.7, 96.5, 92.9, 93.7, 94.7, and 91.2%, respectively. Comparisons were conducted with state-of-the-art (SOTA) techniques, and showed improved accuracy and reduced computational time.