Voice acoustic analysis can be a valuable and objective tool supporting the diagnosis of many neurodegenerative diseases, especially in times of distant medical examination during the pandemic. The article compares the application of selected signal processing methods and machine learning algorithms for the taxonomy of acquired speech signals representing the vowel a with prolonged phonation in patients with Parkinson’s disease and healthy subjects. The study was conducted using three different feature engineering techniques for the generation of speech signal features as well as the deep learning approach based on the processing of images involving spectrograms of different time and frequency resolutions. The research utilized real recordings acquired in the Department of Neurology at the Medical University of Warsaw, Poland. The discriminatory ability of feature vectors was evaluated using the SVM technique. The spectrograms were processed by the popular AlexNet convolutional neural network adopted to the binary classification task according to the strategy of transfer learning. The results of numerical experiments have shown different efficiencies of the examined approaches; however, the sensitivity of the best test based on the selected features proposed with respect to biological grounds of voice articulation reached the value of 97% with the specificity no worse than 93%. The results could be further slightly improved thanks to the combination of the selected deep learning and feature engineering algorithms in one stacked ensemble model.

The self-sealing process, associated with chemical and microstructural changes inside damaged cement-based composites, leads to the recovery of the original material integrity. Assessing the magnitude of internal changes in samples using non-destructive techniques to capture only the self-sealing effects is difficult. The challenge is evaluating the differences between subsequent observations in time and between samples with different properties. This paper proposes a new approach to the use of an ultrasonic technique for self-sealing investigation. The method allows the quantification of material changes strictly related to self-sealing processes, excluding changes caused by the naturally progressing hydration of binders. The applied ultrasonic pulse velocity (UPV) data processing procedure allows the investigation of material changes inside and near the cracks, the effects of stimulating the self-sealing of cement composites with mineral additives, and the assessment of changes over time. An important aspect of the method is the sample preparation procedure and testing conditions that reduce the impact of moisture content on the UPV measurements. New parameters allowing the quantitative characterization of the self-sealing process are proposed. The method was evaluated using cement mortars modified with siliceous fly ash with induced cracks 0 to 750 µm wide, which were then cured in water for 152 days. The maximum degree of effective crack filling as a result of autogenous self-sealing in the tested mortars was determined to range from 33% to 57%. Observations of the microstructure of the crack surface confirmed that apart from the volume of the newly formed products, the density of these products may have a key impact on the ultrasonic measurements of the self-sealing performance. The studies were supplemented by the examination of the compression strength of mortars, mortar sample scanning and computer image processing, and observations using an optical microscope and scanning electron microscope with energy dispersive spectroscopy.

Crack width measurement is an important element of research on the progress of self-healing cement composites. Due to the nature of this research, the method of measuring the width of cracks and their changes over time must meet specific requirements. The article presents a novel method of measuring crack width based on images from a scanner with an optical resolution of 6400 dpi, subject to initial image processing in the ImageJ development environment and further processing and analysis of results. After registering a series of images of the cracks at different times using SIFT conversion (Scale-Invariant Feature Transform), a dense network of line segments is created in all images, intersecting the cracks perpendicular to the local axes. Along these line segments, brightness profiles are extracted, which are the basis for determination of crack width. The distribution and rotation of the line of intersection in a regular layout, automation of transformations, management of images and profiles of brightness, and data analysis to determine the width of cracks and their changes over time are made automatically by own code in the ImageJ and VBA environment. The article describes the method, tests on its properties, sources of measurement uncertainty. It also presents an example of application of the method in research on autogenous self-healing of concrete, specifically the ability to reduce a sample crack width and its full closure within 28 days of the self-healing process.