Wyniki wyszukiwania - BazTech

1

„MIGASYSTENT” – jak sztuczna inteligencja pomaga osobom niesłyszącym

Ciesielski Antoni, Sobczak Paweł

Przegląd Techniczny : Gazeta Inżynierska

|

2025

|

nr 6

18--21

PL

„MigAsystent” to prototyp nowoczesnego urządzenia wspomagającego komunikację zrealizowanego z myślą o osobach niesłyszących oraz instytucjach publicznych. Projekt ten umożliwia tłumaczenie gestów języka migowego [4] (ASL) na mowę i tekst w czasie rzeczywistym, eliminując bariery w kontaktach z osobami niemigającymi. System opiera się na głębokiej sieci neuronowej ResNet-50, kamerze HD, ekranie dotykowym oraz syntezatorze mowy. Rozpoznane gesty są przetwarzane lokalnie, dzięki czemu urządzenie może działać offline, bez konieczności dostępu do sieci Internet. MigAsystent wyróżnia się kompaktową konstrukcją, wielojęzyczną obsługą oraz dostępnością w formie aplikacji na komputery i smartfony. Projekt ten jest przykładem zastosowania sztucznej inteligencji i przetwarzania obrazu w służbie dostępności i integracji społecznej.

EN

“MigAssistant” is a prototype of a modern communication device designed for deaf people and public institutions. It helps translate American Sign Language (ASL) gestures into speech and text in real time. This removes communication barriers between deaf and hearing people. The system uses a deep neural network (ResNet-50), an HD camera, a touch screen, and a speech synthesizer. It works offline, without needing the internet, because all gesture recognition is done locally. MigAssistant is compact, supports many languages, and is also available as an app for computers and smartphones. This project shows how artificial intelligence and image processing can support accessibility and social inclusion.

2

Optimizing Cognitive Radio Networks with Deep Learning-Based Semantic Spectrum Sensing

Mahesh Kumar N., R Arthi

Journal of Telecommunications and Information Technology

|

2024

|

nr 4

78--85

EN

Spectrum aggregation in 4G and 5G networks is a technique used to combine multiple frequency bands to boost communication performance. The cognitive radio feature improves the ability to combine spectrum in LTE and 5G environments by enabling dynamic spectrum sensing. Spectrum sensing is a major problem in spectrum aggregation due to the presence of various types of interference, such as noise. Phase noise is an issue due to its 1 MHz frequency offset experienced within 5G’s 28 GHz operating band, with the distorted signal generating more spectrum sensing-related errors. To solve this problem, the proposed work suggests an optimized deep learning-based semantic spectrum sensing model using three sets of optimizers (ResNet-50, DeepLab V3 and sand cat) offering a high detection accuracy of 99.7%with the optimized training parameter of a high signal-to-noise ratio equaling 40 dB.

3

Development and Optimization of Deep Learning Systems for MRI Analysis in Alzheimer’s Disease Monitoring

Podolszańska Jolanta

Journal of Telecommunications and Information Technology

|

2024

|

nr 4

56--61

EN

Alzheimer’s disease is one of the leading causes of dementia worldwide, and its increasing prevalence presents significant diagnostic and therapeutic challenges, particularly in an aging population. Current diagnostic methods, including patient history reviews, neuropsychological tests, and MRI scans, often fail to achieve adequate sensitivity and specificity levels. In response to these challenges, this study introduces an advanced convolutional neural network (CNN) model that combines ResNet-50 and Inception V3 architectures to classify, with a high degree of precision, the stages of Alzheimer’s disease based on MRI. The model was developed and evaluated using data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and classifies MRI scans into four clinical categories representing different stages of disease severity. The evaluation results, based on the precision, sensitivity and F1 score metrics, demonstrate the effectiveness of the model. Additional augmentation techniques and differential class weighting further enhance the accuracy of the model. Visualization of results using the t-SNE method and the confusion matrix underscores the ability to distinguish between disease categories, supporting the model’s potential to aid in neurological diagnosis and classification.

4

Plant disease detection using ensembled CNN framework

Mondal Subhash, Banerjee Suharta, Mukherjee Subinoy, Sengupta Diganta

Computer Science

|

2022

|

T. 23 (3)

321--333

EN

Agriculture exhibits the prime driving force for the growth of agro-based economies globally. In agriculture, detecting and preventing crops from the attacks of pests is a primary concern in today’s world. The early detection of plant disease becomes necessary in order to avoid the degradation of the yield of crop production. In this paper, we propose an ensemble-based convolutional neural network (CNN) architecture that detects plant disease from the images of a plant’s leaves. The proposed architecture considers CNN architectures like VGG-19, ResNet-50, and InceptionV3 as its base models, and the prediction from these models is used as an input for our meta-model (Inception-ResNetV2). This approach helped us build a generalized model for disease detection with an accuracy of 97.9% under test conditions.