KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Settings

Subject: search.filters.subject.Convolutional neural network (CNN)

Search Results

Now showing 1 - 5 of 5
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Deep Neural Network Detection With an ITI Subtraction for Non-Uniform Track-Width Two-Dimensional Magnetic Recording
    (2024-04-01)
    Buajong, Chaiwat
    ;
    Lee, Jaejin
    ;
    Warisarn, Chanon
    Continuously expanding the magnetic recording density results in unavoidable interferences including intersymbol interference (ISI) and intertrack interference (ITI) that both critically degrade the system performance. Even with advanced signal processing tools, two-dimensional magnetic recording (TDMR) still struggles to provide satisfactory performance. Thus, this article proposes the deep neural network (DNN)-based detection deployed in conjunction with an equalizer for the TDMR system. The coding scheme uses a low-density parity-check (LDPC) code, enabling the information exchange or the turbo decoding. The retrieval of data occurs within a group of three adjacent tracks. We also explore two different track configurations: uniform and non-uniform tracks that involve doubling the width of the middle track among the three adjacent tracks. The utilization of the highly reliable signal obtained from the double-width track enables the application of ITI subtraction technique, enhancing the information exchange. This technique can mitigate the ITI effect by subtracting the target signal with the imitated ITI signal. In addition, we investigate two different DNN architectures including the multilayer perceptron (MLP) and convolutional neural network (CNN), along with two scenarios for the detections in different passes of turbo decoding. The simulation results conducted on the Voronoi media model, with realistic grains and non-magnetic grain boundaries, show that the proposed detection systems with the non-uniform track configuration offer a performance gain up to 5.3 dB over the system with the uniform track configuration. Moreover, iteration for the turbo decoding passes incrementally improves the system performance in the proposed systems with the non-uniform track while the systems with the uniform track no longer provide performance gain as the number of iterations goes on.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Cognitive spectroscopy for the classification of rice varieties: A comparison of machine learning and deep learning approaches in analysing long-wave near-infrared hyperspectral images of brown and milled samples
    (2022-06-01)
    Onmankhong, Jiraporn
    ;
    Ma, Te
    ;
    Inagaki, Tetsuya
    ;
    Sirisomboon, Panmanas
    ;
    Tsuchikawa, Satoru
    Rapid and non-destructive detection of genuine Thai Jasmine rice (Khao Dawk Mali 105 (KDML105)) from Pathum Thani1 (PTT1) and Phitsanulok2 (PSL2) under either milled or brown conditions is required to disrupt fraudulent. This study aimed to resolve this real issue using long-wave near infrared hyperspectral imaging (NIR-HSI) coupled with machine learning and deep learning approaches. The best classification accuracy for the milled rice was achieved using the spectral imaging-based analysis on the NIR-HSI data with selected wavelength, approximately 95% for the test set either by convolutional neural network or support vector machine (SVM), whereas for the brown rice, the SVM model based on the averaged NIR spectra could achieve the best classification accuracy of 95.4%. It suggests the chemical component difference and its spatial distribution in the milled rice could contribute higher classification accuracy. Additionally, the surface bran effects of brown rice could be reduced by using averaged spectral data coupled with the SVM method.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Detection and Classification of COVID-19 Chest X-rays by the Deep Learning Technique
    (2022-01-01)
    Sonarra, Wannika
    ;
    Vongmanee, Naphatsawan
    ;
    Wanluk, Nutthanan
    ;
    Pintavirooj, Chuchat
    ;
    Visitsattapongse, Sarinporn
    The Coronavirus disease (COVID-19) infection has become a pandemic, and this is the most critical problem that has occurred in Thailand and also expanded all over the world. As such, it is not astonishing to know that this virus has had a direct effect on hospitals with the delayed screening of patients because of the increasing number of daily cases and the shortage of medical personnel and restricted treatment space. Due to such restrictions, in this study, we used a clinical decision-making system with predictive algorithms. Predictive algorithms could potentially ease the strain on healthcare systems by identifying the diseases. Moreover, image classification is one interesting aspect of image processing. Convolutional neural network (CNN) is a widely used algorithm for image classification by separating the images of the COVID-19 disease, images with a lung infection, and normal images. To evaluate the predictive performance of our models, precision, F1-score, recall, receiver operating characteristic (ROC) curve (area under the ROC curve), and accuracy scores were used. It was observed that the predictive models trained on the laboratory findings could be used to predict the COVID-19 infection as well and could be helpful for medical experts to appropriately prioritize the resources. This could be employed to assist medical experts in validating their initial laboratory findings and could also be used for clinical prediction studies.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Ionograms Scaling by Using the Convolutional Neural Network
    (2021-03-10)
    Thanakulketsarat, Thananphat
    ;
    Sopon, Thanomsak
    ;
    Phakphisut, Watid
    ;
    Hozumi, Kornyanat
    ;
    Wongtrairat, Wannaree
    Ionosphere in F layer has the most irregularity for phenomenon occurrence of amplitude scintillation which leads to the problem in the satellite signals. Ionosphere can be observed by Ionosonde to study F2 layer critical frequency (foF2) parameter and height of F layer (h'F) parameter from the ionogram. This paper presents the convolutional neural network (CNN) to determine foF2 and h'F parameters. The simulation start from passing the ionogram images to the proposed CNN model with 2,000 epoch training. The simulated accuracies of both foF2 and h'F parameters are equal to 92.8% and 98.4%, respectively.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    ConvXGB: A new deep learning model for classification problems based on CNN and XGBoost
    (2021-02-01)
    Thongsuwan, Setthanun
    ;
    Jaiyen, Saichon
    ;
    Padcharoen, Anantachai
    ;
    Agarwal, Praveen
    We describe a new deep learning model - Convolutional eXtreme Gradient Boosting (ConvXGB) for classification problems based on convolutional neural nets and Chen et al.’s XGBoost. As well as image data, ConvXGB also supports the general classification problems, with a data preprocessing module. ConvXGB consists of several stacked convolutional layers to learn the features of the input and is able to learn features automatically, followed by XGBoost in the last layer for predicting the class labels. The ConvXGB model is simplified by reducing the number of parameters under appropriate conditions, since it is not necessary re-adjust the weight values in a back propagation cycle. Experiments on several data sets from UCL Repository, including images and general data sets, showed that our model handled the classification problems, for all the tested data sets, slightly better than CNN and XGBoost alone and was sometimes significantly better.