Informatica

Proliferation of wearable healthcare devices has created the need to deliver artificial intelligence applications for these resource-constrained devices to achieve faster, localized decision-making, by bringing computation closer to the data sources, for improved responsiveness and privacy. This contribution presents the results of an experimental evaluation of artificial neural network compression techniques, including quantization, structured pruning, and knowledge distillation, applied to multi-label classification of electrocardiogram (ECG) signals. The experiments were carried out on the PTB-XL dataset using three deep learning models, i.e. an LSTM-based recurrent neural network, a 1D convolutional neural network, and a 1D residual neural network. The results show how the compression methods impact model quality and highlight opportunities to reduce model size and accelerate inference, thereby enabling effective deployment on resource-constrained, edge devices.

In this paper the gaze tracking system based on the adaptively changing threshold value of the gray level, which automatically detects the pupil position in two dimensional data is proposed. The system detects closed eye using normalized accumulative luminosity function. The detection of closed eye allows the confirmation of voluntary selected command in a more natural way. The presented technique allows recalibration without intervention or help of other assistive persons. Natural head motion is evaluated, employing signals, generated by orientation sensor. Within two experiments, the proposed system is tested at different behavior modes in respect of speed and accuracy of the system.

In this paper, the main measure, an amount of information, of the information theory is analyzed and corrected. The three conceptions of the theory on the microstate, dissipation pathways, and self‐organization levels with a tight connection to the statistical physics are discussed. The concepts of restricted information were introduced as well as the proof of uniqueness of the entropy function, when the probabilities are rational numbers, is presented.

The development of Lithuanian HMM/ANN speech recognition system, which combines artificial neural networks (ANNs) and hidden Markov models (HMMs), is described in this paper. A hybrid HMM/ANN architecture was applied in the system. In this architecture, a fully connected three‐layer neural network (a multi‐layer perceptron) is trained by conventional stochastic back‐propagation algorithm to estimate the probability of 115 context‐independent phonetic categories and during recognition it is used as a state output probability estimator. The hybrid HMM/ANN speech recognition system based on Mel Frequency Cepstral Coefficients (MFCC) was developed using CSLU Toolkit. The system was tested on the VDU isolated‐word Lithuanian speech corpus and evaluated on a speaker‐independent ∼750 distinct isolated‐word recognition task. The word recognition accuracy obtained was about 86.7%.

An analytical review of recent publications in the area of digital speech signal processing is presented. The aim of the given paper is the analysis of these publications, where Artificial Neural Networks (ANNs) were successfully employed. Numerous methods of ANNs employment are discussed due to identify when and why they are reliable alternative to the conventional adaptive signal processing techniques.