Compact Convolutional Neural Network Research Articles

Subject-Independent Wearable P300 Brain-Computer Interface Based on Convolutional Neural Network and Metric Learning.

The calibration procedure for a wearable P300 brain-computer interface (BCI) greatly impact the user experience of the system. Each user needs to spend additional time establishing a decoder adapted to their own brainwaves. Therefore, achieving subject independent is an urgent issue for wearable P300 BCI needs to be addressed. A dataset of electroencephalogram (EEG) signals was constructed from 100 individuals by conducting a P300 speller task with a wearable EEG amplifier. A framework is proposed that initially improves cross-subject consistency of EEG features through a common feature extractor. Subsequently, a simple and compact convolutional neural network (CNN) architecture is employed to learn an embedding sub-space, where the mapped EEG features are maximally separated, while pursuing the minimum distance within the same class and the maximum distance between different classes. Finally, the model's generalization capability was further optimized through fine-tuning. The proposed method significantly boosts the average accuracy of wearable P300 BCI to 73.23 ±7.62% without calibration and 78.75±6.37% with fine-tuning. The results demonstrate the feasibility and excellent performance of our dataset and framework. A calibration-free wearable P300 BCI system is feasible, suggesting significant potential for practical applications of the wearable P300 BCI system.

Skin cancer classification leveraging multi-directional compact convolutional neural network ensembles and gabor wavelets

Skin cancer (SC) is an important medical condition that necessitates prompt identification to ensure timely treatment. Although visual evaluation by dermatologists is considered the most reliable method, its efficacy is subjective and laborious. Deep learning-based computer-aided diagnostic (CAD) platforms have become valuable tools for supporting dermatologists. Nevertheless, current CAD tools frequently depend on Convolutional Neural Networks (CNNs) with huge amounts of deep layers and hyperparameters, single CNN model methodologies, large feature space, and exclusively utilise spatial image information, which restricts their effectiveness. This study presents SCaLiNG, an innovative CAD tool specifically developed to address and surpass these constraints. SCaLiNG leverages a collection of three compact CNNs and Gabor Wavelets (GW) to acquire a comprehensive feature vector consisting of spatial–textural–frequency attributes. SCaLiNG gathers a wide range of image details by breaking down these photos into multiple directional sub-bands using GW, and then learning several CNNs using those sub-bands and the original picture. SCaLiNG also combines attributes taken from various CNNs trained with the actual images and subbands derived from GW. This fusion process correspondingly improves diagnostic accuracy due to the thorough representation of attributes. Furthermore, SCaLiNG applies a feature selection approach which further enhances the model’s performance by choosing the most distinguishing features. Experimental findings indicate that SCaLiNG maintains a classification accuracy of 0.9170 in categorising SC subcategories, surpassing conventional single-CNN models. The outstanding performance of SCaLiNG underlines its ability to aid dermatologists in swiftly and precisely recognising and classifying SC, thereby enhancing patient outcomes.

A Novel Approach for Meat Quality Assessment Using an Ensemble of Compact Convolutional Neural Networks

The rising awareness of nutritional values has led to an increase in the popularity of meat-based diets. Hence, to ensure public health, industries and consumers are focusing more on the quality and freshness of this food. Authentic meat quality assessment methods can indeed be exorbitant and catastrophic. Furthermore, it is subjective and reliant on the knowledge of specialists. Fully automated computer-aided diagnosis systems are required to eradicate the variability among experts. However, evaluating the quality of meat automatically is challenging. Deep convolutional neural networks have shown a tremendous improvement in meat quality assessment. This research intends to utilize an ensemble framework of shallow convolutional neural networks for assessing the quality and freshness of the meat. Two compact CNN architectures (ConvNet-18 and ConvNet-24) are developed, and the efficacy of the models are evaluated using two publicly available databases. Experimental findings reveal that the ConvNet-18 outperforms other state-of-the models in classifying fresh and spoiled meat with an overall accuracy of 99.4%, whereas ConvNet-24 shows a better outcome in categorizing the meat based on its freshness. This model yields an accuracy of 96.6%, which is much better compared with standard models. Furthermore, the suggested models effectively detect the quality and freshness of the meat with less complexity than the existing state-of-the art techniques.

Text-independent voiceprint recognition via compact embedding of dilated deep convolutional neural networks

In order to process speech, most state-of-the-art experimental methods employ convolutional neural networks (CNNs), which operate on a continuous, 1-dimensional (1-D) time stream. In an audio signal, the mel-spectrogram facilitates the representation of attributes of the utterances' in the frequency domain (which corresponds to the speech spectrum). Moreover, for a time-series speaker signal, CNNs are superior to machine or transfer learning models in capturing characteristics from long-form talks. This paper introduces a jump-connected 1-D CNN that employs a combined loss function for speaker recognition. The suggested model uses a 1-D convolutional layer combined with jump connections to extract speaker-specific characteristics; this reduces time-based and frequency-based variability for faster computing. A combined softmax loss, stable L2-norm, and smooth L1-norm loss function guide the proposed compact convolutional neural networks (CCNN) to identify the correct spokesman with improved efficacy. We evaluated the proposed framework using various standard and real-time audio datasets. The experimental findings demonstrate that the proposed CCNN outperforms existing approaches by reducing the equal error rate by 9.02 %. Also, our recommended voiceprint identification model achieves an impressive average speaker recognition rate of 98.76 %. Simultaneously, the reliability of the 1-D CCNN is tested under various conditions. Other fields of study, like language modelling, could employ this approach after some fine-tuning.Relevance of the work: Speaker recognition is an area of interest in which machine learning (ML) and deep learning (DL) schemes, when combined, have the potential to make history in the areas of forensic sciences, automation, and authentication. Using a modest CNN can enhance the identification and verification process by ignoring many issues such as false positives, background noise, and so on. Expanding this process would facilitate raga identification and disease treatment therapies.

An end-to-end multi-task motor imagery EEG classification neural network based on dynamic fusion of spectral-temporal features

Electroencephalograph (EEG) brain-computer interfaces (BCI) have potential to provide new paradigms for controlling computers and devices. The accuracy of brain pattern classification in EEG BCI is directly affected by the quality of features extracted from EEG signals. Currently, feature extraction heavily relies on prior knowledge to engineer features (for example from specific frequency bands); therefore, better extraction of EEG features is an important research direction. In this work, we propose an end-to-end deep neural network that automatically finds and combines features for motor imagery (MI) based EEG BCI with 4 or more imagery classes (multi-task). First, spectral domain features of EEG signals are learned by compact convolutional neural network (CCNN) layers. Then, gated recurrent unit (GRU) neural network layers automatically learn temporal patterns. Lastly, an attention mechanism dynamically combines (across EEG channels) the extracted spectral-temporal features, reducing redundancy. We test our method using BCI Competition IV-2a and a data set we collected. The average classification accuracy on 4-class BCI Competition IV-2a was 85.1 % ± 6.19 %, comparable to recent work in the field and showing low variability among participants; average classification accuracy on our 6-class data was 64.4 % ± 8.35 %. Our dynamic fusion of spectral-temporal features is end-to-end and has relatively few network parameters, and the experimental results show its effectiveness and potential.

Ultra low-power, wearable, accelerated shallow-learning fall detection for elderly at-risk persons

This work focuses on the development and manufacturing of a wireless, wearable, low-power fall detection sensor (FDS) designed to predict and detect falls in elderly at-risk individuals. Unintentional falls are a significant risk in this demographic, often resulting from diminished physical capabilities such as reduced hand grip strength and complications from conditions like arthritis, vertigo, and neuromuscular issues. To address this, we utilize advanced low-power field-programmable gate arrays (FPGAs) to implement a fixed-function neural network capable of categorizing activities of daily life (ADLs), including the detection of falls. This system employs a Convolutional Neural Network (CNN) model, trained and validated using the Caffe deep learning framework with data collected from human subjects experiments. This system integrates an ST Microelectronics LSM6DSOX inertial measurement unit (IMU) sensor, embedded with an ultra-low-power Lattice iCE40UP FPGA, which samples and stores joint acceleration and orientation rate. Additionally, we have acquired and published a dataset of 3D accelerometer and gyroscope measurements from predefined ADLs and falls, using volunteer human subjects. This innovative approach aims to enhance the safety and well-being of older adults by providing timely and accurate fall detection and prediction.In this paper, we present an innovative approach to utilizing a compact Convolutional Neural Network (CNN) core for accelerating convolutional operations on a machine learning model, suitable for deployment on an ultra-low power FPGA.

A Case Study of a Tiny Machine Learning Application for Battery State-of-Charge Estimation

Growing battery use in energy storage and automotive industries demands advanced Battery Management Systems (BMSs) to estimate key parameters like the State of Charge (SoC) which are not directly measurable using standard sensors. Consequently, various model-based and data-driven approaches have been developed for their estimation. Among these, the latter are often favored due to their high accuracy, low energy consumption, and ease of implementation on the cloud or Internet of Things (IoT) devices. This research focuses on creating small, efficient data-driven SoC estimation models for integration into IoT devices, specifically the Infineon Cypress CY8CPROTO-062S3-4343W. The development process involved training a compact Convolutional Neural Network (CNN) and an Artificial Neural Network (ANN) offline using a comprehensive dataset obtained from five different batteries. Before deployment on the target device, model quantization was performed using Infineon’s ModusToolBox Machine Learning (MTB-ML) configurator 2.0 software. The tests show satisfactory results for both chosen models with a good accuracy achieved, especially in the early stages of the battery lifecycle. In terms of the computational burden, the ANN has a clear advantage over the more complex CNN model.

Vision-aided grasp classification: design and evaluation of compact CNN for prosthetic hands

Powered prosthetic hands capable of executing various grasp patterns are highly sought-after solutions for upper limb amputees. A crucial requirement for such prosthetic hands is the accurate identification of the intended grasp pattern and subsequent activation of the prosthetic digits accordingly. Vision-based grasp classification techniques offer improved coordination between amputees and prosthetic hands without physical contact. Deep learning methods, particularly Convolutional Neural Networks (CNNs), are utilized to process visual information for classification. The key challenge lies in developing a model that can effectively generalize across various object shapes and accurately classify grasp classes. To address this, a compact CNN model named GraspCNet is proposed, specifically designed for grasp classification in prosthetic hands. The use of separable convolutions reduces the computational burden, making it potentially suitable for real-time applications on embedded systems. The GraspCNet model is designed to learn and generalize from object shapes, allowing it to effectively classify unseen objects beyond those included in the training dataset. The proposed model was trained and tested using various standard object data sets. A cross-validation strategy has been adopted to perform better in seen and unseen object class scenarios. The average accuracy achieved was 82.22% and 75.48% in the case of seen, and unseen object classes respectively. In computer-based real-time experiments, the GraspCNet model achieved an accuracy of 69%. A comparative analysis with state-of-the-art techniques revealed that the proposed GraspCNet model outperformed most benchmark techniques and demonstrated comparable performance with the DcnnGrasp method. The compact nature of the GraspCNet model suggests its potential for integration with other sensing modalities in prosthetic hands.

End-to-end automated speech recognition using a character based small scale transformer architecture

This study explores the feasibility of constructing a small-scale speech recognition system capable of competing with larger, modern automated speech recognition (ASR) systems in both performance and word error rate (WER). Our central hypothesis posits that a compact transformer-based ASR model can yield comparable results, specifically in terms of WER, to traditional ASR models while challenging contemporary ASR systems that boast significantly larger computational sizes. The aim is to extend ASR capabilities to under-resourced languages with limited corpora, catering to scenarios where practitioners face constraints in both data availability and computational resources. The model, comprising a compact convolutional neural network (CNN) and transformer architecture with 2.214 million parameters, challenges the conventional wisdom that large-scale transformer-based ASR systems are essential for achieving high accuracy. In comparison, contemporary ASR systems often deploy over 300 million parameters. Trained on a modest dataset of approximately 3000 h – significantly less than the 50,000 h used in larger systems – the proposed model leverages the Common Voice and LibriSpeech datasets. Evaluation on the LibriSpeech test-clean and test-other datasets produced character error rates (CERs) of 6.40% and 16.73% and WERs of 16.03% and 35.51% respectively. Comparisons with existing architectures showcase the efficiency of our model. A gated recurrent unit (GRU) architecture, albeit achieving lower error rates, incurred a computational cost 24 times larger than our proposed model. Large-scale transformer architectures, while achieving marginally lower WERs (2%–4% on LibriSpeech test-clean), require 200 times more parameters and 53,000 additional hours of training data. Modern large language models are used to improve the WERs, but require large computational resources. To further enhance performance, a small 4-g language model was integrated into our end-to-end ASR model, resulting in improved WERs. The overarching goal of this work is to provide a practical solution for practitioners dealing with limited datasets and computational resources, particularly in the context of under-resourced languages.

Bat2Web: A Framework for Real-Time Classification of Bat Species Echolocation Signals Using Audio Sensor Data.

Bats play a pivotal role in maintaining ecological balance, and studying their behaviors offers vital insights into environmental health and aids in conservation efforts. Determining the presence of various bat species in an environment is essential for many bat studies. Specialized audio sensors can be used to record bat echolocation calls that can then be used to identify bat species. However, the complexity of bat calls presents a significant challenge, necessitating expert analysis and extensive time for accurate interpretation. Recent advances in neural networks can help identify bat species automatically from their echolocation calls. Such neural networks can be integrated into a complete end-to-end system that leverages recent internet of things (IoT) technologies with long-range, low-powered communication protocols to implement automated acoustical monitoring. This paper presents the design and implementation of such a system that uses a tiny neural network for interpreting sensor data derived from bat echolocation signals. A highly compact convolutional neural network (CNN) model was developed that demonstrated excellent performance in bat species identification, achieving an F1-score of 0.9578 and an accuracy rate of 97.5%. The neural network was deployed, and its performance was evaluated on various alternative edge devices, including the NVIDIA Jetson Nano and Google Coral.

An Efficient Hybrid Deep Learning Accelerator for Compact and Heterogeneous CNNs

Resource-efficient Convolutional Neural Networks (CNNs) are gaining more attention. These CNNs have relatively low computational and memory requirements. A common denominator among such CNNs is having more heterogeneity than traditional CNNs. This heterogeneity is present at two levels: intra-layer type and inter-layer type. Generic accelerators do not capture these levels of heterogeneity, which harms their efficiency. Consequently, researchers have proposed model-specific accelerators with dedicated engines. When designing an accelerator with dedicated engines, one option is to dedicate one engine per CNN layer. We refer to accelerators designed with this approach as single-engine single-layer (SESL). This approach enables optimizing each engine for its specific layer. However, such accelerators are resource-demanding and unscalable. Another option is to design a minimal number of dedicated engines such that each engine handles all layers of one type. We refer to these accelerators as single-engine multiple-layer (SEML). SEML accelerators capture the inter-layer-type but not the intra-layer-type heterogeneity. We propose the Fixed Budget Hybrid CNN Accelerator (FiBHA), a hybrid accelerator composed of an SESL part and an SEML part, each processing a subset of CNN layers. FiBHA captures more heterogeneity than SEML while being more resource-aware and scalable than SESL. Moreover, we propose a novel module, Fused Inverted Residual Bottleneck (FIRB), a fine-grained and memory-light SESL architecture building block. The proposed architecture is implemented and evaluated using high-level synthesis (HLS) on different Field Programmable Gate Arrays representing various resource budgets. Our evaluation shows that FiBHA improves the throughput by up to 4 x and 2.5 x compared to state-of-the-art SESL and SEML accelerators, respectively. Moreover, FiBHA reduces memory and energy consumption compared to an SEML accelerator. The evaluation also shows that FIRB reduces the required memory by up to 54%, and energy requirements by up to 35% compared to traditional pipelining.

MixedSCNet: LiDAR-Based Place Recognition Using Multi-Channel Scan Context Neural Network

In the realm of LiDAR-based place recognition tasks, three predominant methodologies have emerged: manually crafted feature descriptor-based methods, deep learning-based methods, and hybrid methods that combine the former two. Manually crafted feature descriptors often falter in reverse visits and confined indoor environments, while deep learning-based methods exhibit limitations in terms of generalization to distinct data domains. Hybrid methods tend to fix these problems, albeit at the cost of an expensive computational burden. In response to this, this paper introduces MixedSCNet, a novel hybrid approach designed to harness the strengths of manually crafted feature descriptors and deep learning models while keeping a relatively low computing overhead. MixedSCNet starts with constructing a BEV descriptor called MixedSC, which takes height, intensity, and smoothness into consideration simultaneously, thus offering a more comprehensive representation of the point cloud. Subsequently, MixedSC is fed into a compact Convolutional Neural Network (CNN), which further extracts high-level features, ultimately yielding a discriminative global point cloud descriptor. This descriptor is then employed for place retrieval, effectively bridging the gap between manually crafted feature descriptors and deep learning models. To substantiate the efficacy of this amalgamation, we undertake an extensive array of experiments on the KITTI and NCLT datasets. Results show that MixedSCNet stands out as the sole method showcasing state-of-the-art performance across both datasets, outperforming the other five methods while maintaining a relatively short runtime.

SparseVoxNet: 3-D Object Recognition With Sparsely Aggregation of 3-D Dense Blocks.

Automatic recognition of 3-D objects in a 3-D model by convolutional neural network (CNN) methods has been successfully applied to various tasks, e.g., robotics and augmented reality. Three-dimensional object recognition is mainly performed by analyzing the object using multi-view images, depth images, graphs, or volumetric data. In some cases, using volumetric data provides the most promising results. However, existing recognition techniques on volumetric data have many drawbacks, such as losing object details on converting points to voxels and the large size of the input volume data that leads to substantial 3-D CNNs. Using point clouds could also provide very promising results; however, point-cloud-based methods typically need sparse data entry and time-consuming training stages. Thus, using volumetric could be a more efficient and flexible recognizer for our special case in the School of Medicine, Shanghai Jiao Tong University. In this article, we propose a novel solution to 3-D object recognition from volumetric data using a combination of three compact CNN models, low-cost SparseNet, and feature representation technique. We achieve an optimized network by estimating extra geometrical information comprising the surface normal and curvature into two separated neural networks. These two models provide supplementary information to each voxel data that consequently improve the results. The primary network model takes advantage of all the predicted features and uses these features in Random Forest (RF) for recognition purposes. Our method outperforms other methods in training speed in our experiments and provides an accurate result as good as the state-of-the-art.

Özel Derin Konvolüsyonel Sinir Ağı Kullanarak Cilt Kanseri Tanıma

Skin cancer is a common form of cancer that affects millions of people worldwide. Early detection and accurate diagnosis of skin cancer are crucial for effective treatment and management of the disease. There has been a growing interest in using deep learning techniques and computer vision algorithms to develop automated skin cancer detection systems in recent years. Among these techniques, convolutional neural networks (CNN) have shown remarkable performance in detecting and classifying skin lesions. This paper presents a comprehensive study using CNN and deep learning techniques for skin cancer detection using the International Skin Imaging Collaboration (ISIC) dataset. The proposed architecture is a compact deep CNN that is trained using a dataset of benign and malignant skin lesion images. The proposed architecture has achieved 84.8% accuracy, 83.8% TPR, 83.7% TNR, 81.6% F1-score and 80.5% precision for performance evaluation. The experimental results show promising results for the accurate and efficient detection of skin cancer, which has the potential to improve the diagnosis and treatment of this life-threatening disease.

Classification of Epileptic Seizure Types Using Multiscale Convolutional Neural Network and Long Short-Term Memory

The accurate classification of seizure types using electroencephalography (EEG) signals plays a vital role in determining a precise treatment plan and therapy for epilepsy patients. Among the available deep network models, Convolutional Neural Networks (CNNs) are the most widely adopted models for learning and representing EEG signals. However, typical CNNs have high computational complexity, leading to overfitting problems. This paper proposes the design of two effective, lightweight deep network models; the 1D multiscale neural network (1D-MSCNet) model and the Long Short-term Memory (LSTM)-based compact CNN (EEG-LSTMNet) model. The 1D-MSCNet model comprises three modules: a spectral–temporal convolution module, a spatial convolution module, and a classification module. It extracts features from input EEG trials at multiple frequency/time ranges, identifying relationships between the spatial distribution of their channels. The EEG-LSTMNet model includes three convolutional layers, namely temporal, depthwise, and separable layers, a single LSTM layer, and two fully connected classification layers to extract discriminative EEG feature representations. Both models have been applied to the same EEG trials collected from the Temple University Hospital (TUH) database. Results revealed F1-score values of 96.9% and 98.4% for the 1D-MSCNet and EEG-LSTMNet, respectively. Based on the demonstrated outcomes, both models outperform related state-of-the-art methods due to their architectures’ adoption of 1D modules and layers that reduce the computational effort needed, solve the overfitting problem, and enhance classification efficiency. Hence, both models could be valuable additions for neurologists to help them decide upon precise treatments and drugs for patients depending on their type of seizure.

Efficient Microalgae Species Identification using Compact Convolutional Neural Network

In this study, we propose a novel approach for microscopic algae species classification by implementing a compact Convolutional Neural Network (CNN) model. Our methodology was tested on a diverse dataset consisting of 18 distinct species of microscopic algae, demonstrating a remarkable classification accuracy exceeding 99%. The outstanding performance of this model is attributed to its compact architecture which maintains high precision while minimizing computational resources, making it a feasible option for real-time applications. Furthermore, we incorporated advanced data augmentation techniques to enhance the generalization capability of our model. By artificially expanding the training dataset, we effectively increased the model's robustness to variance in input data, which significantly contributed to the model's high classification accuracy. The research findings underscore the potential of compact CNN models coupled with data augmentation strategies in high-precision microscopic algae classification tasks, paving the way for future innovations in the field of aquatic microbiology and environmental monitoring.

Temporal-spatial convolutional residual network for decoding attempted movement related EEG signals of subjects with spinal cord injury

Brain Computer Interface (BCI) offers a promising approach to restoring hand functionality for people with cervical spinal cord injury (SCI). A reliable classification of brain activities based on appropriate flexibility in feature extraction could enhance BCI systems performance. In the present study, based on convolutional layers with temporal-spatial, Separable and Depthwise structures, we develop Temporal-Spatial Convolutional Residual Network)TSCR-Net(and Temporal-Spatial Convolutional Iterative Residual Network)TSCIR-Net(structures to classify electroencephalogram (EEG) signals. Using EEG signals in five different hand movement classes of SCI people, we compare the effectiveness of TSCIR-Net and TSCR-Net models with some competitive methods. We use the bayesian hyperparameter optimization algorithm to tune the hyperparameters of compact convolutional neural networks. In order to show the high generalizability of the proposed models, we compare the results of the models in different frequency ranges. Our proposed models decoded distinctive characteristics of different movement efforts and obtained higher classification accuracy than previous deep neural networks. Our findings indicate that TSCIR-Net and TSCR-Net models fulfills a better classification accuracy of 71.11%, and 64.55% for EEG_All and 57.74%, and 67.87% for EEG_Low frequency data sets than the compared methods in the literature.

JPEG2000-Based Semantic Image Compression using CNN

Some of the computer vision applications such as understanding, recognition as well as image processing are some areas where AI techniques like convolutional neural network (CNN) have attained great success. AI techniques are not very frequently used in applications like image compression which are a part of low-level vision applications. Intensifying the visual quality of the lossy video/image compression has been a huge obstacle for a very long time. Image processing tasks and image recognition can be addressed with the application of deep learning CNNs as a result of the availability of large training datasets and the recent advances in computing power. This paper consists of a CNN-based novel compression framework comprising of Compact CNN (ComCNN) and Reconstruction CNN (RecCNN) where they are trained concurrently and ideally consolidated into a compression framework, along with MS-ROI (Multi Structure-Region of Interest) mapping which highlights the semiotically notable portions of the image. The framework attains a mean PSNR value of 32.9dB, achieving a gain of 3.52dB and attains mean SSIM value of 0.9262, achieving a gain of 0.0723dB over the other methods when compared using the 6 main test images. Experimental results in the proposed study validate that the architecture substantially surpasses image compression frameworks, that utilized deblocking or denoising post- processing techniques, classified utilizing Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index Measures (SSIM) with a mean PSNR, SSIM and Compression Ratio of 38.45, 0.9602 and 1.75x respectively for the 50 test images, thus obtaining state-of-art performance for Quality Factor (QF)=5.

A compact CNN model for automated detection of COVID-19 using thorax x-ray images

COVID-19 is an epidemic, causing an enormous death toll. The mutational changing of an RNA virus is causing diagnostic complexities. RT-PCR and Rapid Tests are used for the diagnosis, but unfortunately, these methods are ineffective in diagnosing all strains of COVID-19. There is an utmost need to develop a diagnostic procedure for timely identification. In the proposed work, we come up with a lightweight algorithm based on deep learning to develop a rapid detection system for COVID-19 with thorax chest x-ray (CXR) images. This research aims to develop a fine-tuned convolutional neural network (CNN) model using improved EfficientNetB5. Design is based on compound scaling and trained on the best possible feature extraction algorithm. The low convergence rate of the proposed work can be easily deployed into limited computational resources. It will be helpful for the rapid triaging of victims. 2-fold cross-validation further improves the performance. The algorithm proposed is trained, validated, and testing is performed in the form of internal and external validation on a self-collected and compiled a real-time dataset of CXR. The training dataset is relatively extensive compared to the existing ones. The performance of the proposed technique is measured, validated, and compared with other state-of-the-art pre-trained models. The proposed methodology gives remarkable accuracy (99.5%) and recall (99.5%) for biclassification. The external validation using two different test dataset also give exceptional predictions. The visual depiction of predictions is represented by Grad-CAM maps, presenting the extracted features of the predicted results.

Portable deep-learning decoder for motor imaginary EEG signals based on a novel compact convolutional neural network incorporating spatial-attention mechanism.

Due to high computational requirements, deep-learning decoders for motor imaginary (MI) electroencephalography (EEG) signals are usually implemented on bulky and heavy computing devices that are inconvenient for physical actions. To date, the application of deep-learning techniques in independent portable brain-computer-interface (BCI) devices has not been extensively explored. In this study, we proposed a high-accuracy MI EEG decoder by incorporating spatial-attention mechanism into convolution neural network (CNN), and deployed it on fully integrated single-chip microcontroller unit (MCU). After the CNN model was trained on workstation computer using GigaDB MI datasets (52 subjects), its parameters were then extracted and converted to build deep-learning architecture interpreter on MCU. For comparison, EEG-Inception model was also trained using the same dataset, and was deployed on MCU. The results indicate that our deep-learning model can independently decode imaginary left-/right-hand motions. The mean accuracy of the proposed compact CNN reaches 96.75 ± 2.41% (8 channels: Frontocentral3 (FC3), FC4, Central1 (C1), C2, Central-Parietal1 (CP1), CP2, C3, and C4), versus 76.96 ± 19.08% of EEG-Inception (6 channels: FC3, FC4, C1, C2, CP1, and CP2). To the best of our knowledge, this is the first portable deep-learning decoder for MI EEG signals. The findings demonstrate high-accuracy deep-learning decoding of MI EEG in a portable mode, which has great implications for hand-disabled patients. Our portable system can be used for developing artificial-intelligent wearable BCI devices, as it is less computationally expensive and convenient for real-life application.