Search

This paper discusses the option of hardware-software integration, according to the assessment of human cognitive potential. The presented complex is based on an assessment of the integrative activity of the brain that allows one to get an idea of the current state of relations between the different areas of the brain structures (i.e., allows numerically estimate the number of structures involved in the activity). Evaluation of the integrative activity performed by a coherent analysis of the patient's EEG. The analysis is performed in real time using module coherent analysis developed in OpenVibe software.

In paper different standards for intelligent medical systems and multimodal architecture interfaces are under review. Application of Fast Healthcare Interoperability Resource (FHIR) standard for data model of intelligent multimodal interface are proposed. Intelligent multimodal interface for EEG and accelerometer bio-signals based on FHIR are shown. Use case diagrams, resource diagrams and class diagrams for production and research mode of this interface are described.

Neural networks applied in control loops and safety-critical domains have to meet more requirements than just the overall best function approximation. On the one hand, a small approximation error is required; on the other hand, the smoothness and the monotonicity of selected input–output relations have to be guaranteed. Otherwise, the stability of most of the control laws is lost. In this article we compare two neural network-based approaches incorporating partial monotonicity by structure, namely the Monotonic Multi-Layer Perceptron (MONMLP) network and the Monotonic MIN–MAX (MONMM) network. We show the universal approximation capabilities of both types of network for partially monotone functions. On a number of datasets, we investigate the advantages and disadvantages of these approaches related to approximation performance, training of the model and convergence.

Neural networks applied in control loops and safety-critical domains have to meet more requirements than just the overall best function approximation. On the one hand, a small approximation error is required, on the other hand, the smoothness and the monotonicity of selected input-output relations have to be guaranteed. Otherwise the stability of most of the control laws is lost. Three approaches for partially monotonic models are compared in this article, namely Bounded Derivative Network (BDN) [1], Monotonic Multi-Layer Perceptron Network (MONMLP) [2], and Constrained Linear Regression (CLR). Authors investigated the advantages and disadvantages of these approaches related to approximation performance, training of the model and convergence.