https://etalpykla.vilniustech.lt/handle/123456789/159398 2026-04-11T21:28:04Z https://etalpykla.vilniustech.lt/handle/123456789/159599 Fusion of Activity Recognition and Recurrent Neural Network for Attitude Estimation Improvement Gedminas, Dovydas; Dumpis, Martynas; Serackis, Artūras The paper focuses on machine learning-based processing of inertial measurement unit signals for attitude estimation. Signals from the accelerometer, gyroscope, and magnetometer are used as input to the trained machine learning models, based on the recurrent neural network. Models provides four quaternion parameters predicted by a pre-trained neural network. The practical application of such a system showed that it is hard to get a universal model that is suitable for precise attitude estimation on different types of activity. In this paper, a two-step solution is proposed, constructed from an activity recognition stage and switchable models for the prediction of quaternion parameters followed by attitude estimation. An experimental investigation was performed on publicly available data taken from the Berlin Robust Orientation Estimation Assessment Dataset. The tests, carried out with labeled data, showed that the preparation of activity-related quaternion parameter prediction models can decrease the mean error in attitude estimation by 12.5 % together with a reduction in standard deviation by 3.2 %. 2022-01-01T00:00:00Z https://etalpykla.vilniustech.lt/handle/123456789/159598 Inertial Sensor based System for Upper Limb Motion Quantification Dumpis, Martynas; Gedminas, Dovydas; Serackis, Artūras The paper focuses on the quantification of upper limb motion during human-operated exercises, dedicated to rehabilitation. Continuous tracking of a measurement of each exercise helps the patient individually monitor the current progress of rehabilitation treatment and provide a source of quantified data for the preparation of biological feedback. While the camera-based solutions may provide more precise upper limb tracking, such systems may cause some privacy issues or might be uncomfortable for the patient to use in daily manner. The sensor-based solution is a good option here. Inertial sensors In this investigation, an algorithm for automated calibration of a single sensor with precise estimation of upper limb movement is presented. The study focuses on the measurement of rehabilitation exercises and the collection of parameters that are important to medical diagnosis and long-term monitoring of the change in these parameters. The suggested solution simplifies the human motion estimation process by correctly estimating human motion even inertial sensor's axis is not aligned correctly with the body frame. Additionally, a system with two sensors is proposed to compensate for upper limb movement quantification uncertainty during specific exercises. 2022-01-01T00:00:00Z https://etalpykla.vilniustech.lt/handle/123456789/159597 Recognition of Explosive Objects Using Computer Vision and Machine Learning Mordyk, Oleksandr This article considers an approach to the recognition of explosive objects using a custom object detection model with Tensor-flow framework and OpenCV. The approach to creating a customer's own SSD model is considered in detail. Analyzed the benefits of using OpenCV to deploy an explosive object detection system. Briefly describe the application for testing and visualizing the work of the resulting model. The purpose of the research is using machine learning and computer vision as a new approach for resolving problem of detecting explosive objects. The object of research - the process of detecting explosive objects. Methods of research - methods of object detection, methods of machine learning, methods of simulation. 2022-01-01T00:00:00Z https://etalpykla.vilniustech.lt/handle/123456789/159596 Investigation of Pneumonia Detection using Convolutional Neural Networks Cicėnas, Benediktas; Abromavičius, Vytautas With the advance of Covid-19 pneumonia and its complications increased the load of radiologists several times. The pneumonia detection usually is performed by examination of chest X-Ray radiograph by the radiologists. This process is tedious, influenced by fatigue and may lead to mistakes. Computer-aided diagnostic systems shows the potential for improving accuracy. In this work, we present investigation for pneumonia detection based on several convolutional neural network architectures via transfer learning. Additionally, we propose a method for preparing data from the Radiological Society of North America Pneumonia Detection Challenge to achieve better results. 2022-01-01T00:00:00Z