|
|
 Выпуск 66
- Bessonov M., Pasha Farhadov M. Algorithms for interpretation of prosodic features in low-bitrate speech processing
We study the language identification problem using prosodic features. Prosodic features such as melody, rhythm, timbre and others are difficult to formalize mathematically. Two algorithms for a complex description of prosodic features are proposed in the paper. The first is based on the broad phonetic categories, and the second is based on the cross-correlation of the speech melody and the short-term energy sequence. The fundamental frequency was estimated by MELP algorithm. The performance of the proposed algorithms was evaluated experimentally on a database of speech recordings obtained from Internet and therefore encoded by low-bitrate vocoders. The database includes ten different languages. The proposed algorithms provide a feature description and a multi-layer neural network was used as a language classifier. Both algorithms show satisfactory classification performance, but the broad phonetic categories approach performs slightly better than the cross-correlation function. These algorithms can be applied to a speech signal processed by low-bitrate vocoders without decoding to the original signal.
- Vykhovanets V. Information system using conceptual domain model
Conceptual modeling is used for formal specification of domain knowledge and its integration into an information system. We describe an intelligent information system based on the conceptual domain modeling. The system consists of four layers: client, knowledge, logic and data. The knowledge layer is based on a conceptual model which consists of a conceptual structure and a content of concepts. A conceptual structure is specified as a set of concepts, which contains four abstractions: generalization, typification, association and aggregation. A content of each concept is specified by database tables. The significant difference between the proposed conceptual model and others is the description of the association not as a relationship but as an ordinary concept. This implies that the conceptual structure can be represented as a tree. And the information needed for knowledge-based inference is contained in the concept structure itself. We show that the semantic invariance of the conceptual interpretation improves technological and operational characteristics of the information system.
- Kuznetsov A. Allocation of limited resources in a system with a stable hierarchy (on the example of prospective military communications system)
The article is devoted to the resource allocation problem in a hierarchical social system where resources are distributed according to an agent's role in the system. We propose a general mathematical formulation and apply it to the problem of channel allocation in a special-purpose communication network. Each channel belongs to a certain class and the agents’ hierarchy impose constraints on the allocation such that a pair of agents with given roles should be assigned with channels from given classes. These constraints are represented as a structure which is called needs graph. An algorithm of the automatic partitioning of a radio communication system in accordance with the roles of the nodes is proposed. None of the nodes have complete knowledge about the entire communication system so an information about roles of the nodes is unknown in advance and is established by a beacon exchange process. We illustrate the proposed approach by an example of a special-purpose military communication networks
- Gorodetsky V., Bukhvalov O., Skobelev P., Mayorov I. Industrial applications of multi-agent systems: current state and prospects
Multi-agent system are perceived as one of the most promising technologies for distributed systems development. However current evidences show a lack of industry-level applications. The paper examines the main trends and prospects in development of industrial applications of multi-agent systems and technology, as well as analyzes the recent forecasts and the current state-of-the art with its practical applications. A survey of currently present researches, techniques, frameworks and standards is provided and supported with the most recent statistical data. The paper also analyzes external and internal obstacles for broad commercialization of multi-agent systems and technologies and the lessons learnt through these analyses. The paper describes properties and types of current and future applications for which multi-agent technology has unquestionable advantages including enterprise management, aviation and space, energetics, transportation etc. The paper also shows that multi-agent systems and technologies currently have no alternatives in self-organized control of large-scale objects of network structure.
- Miroshnik S., Gonchar D., Furugyan M. Real-time database structure optimization
The problem of real-time database structure optimization is considered. The objective is to minimize the information redundancy with respect to real-time usage. There is a set of active processes, each process uses several data fields from a given database. The fields should be divided into a set of files such that each process need not use more than one file for normal operation. Different measures are proposed for intra-file, inter-file and inter-process redundancies. We focus on the problem of inter- and intra-file redundancy minimization which can be reduced to a Boolean programming problem. We propose two algorithms for the database structure optimization problem. The first algorithm is based on the Boolean programming reduction and the second one is heuristics with a polynomial computational complexity. An illustrative example is provided. The algorithms are included into a CAD for real-time systems which can be used for flight experiments, ecological monitoring and other fields.
- Vedeshenkov V. A fragmentation approach to diagnosis of digital systems components with minimal quasicomplete graph structures (an example of 7x7 graph)
The problem of fault diagnosis in a multi-processor computing system is considered. The model of a multi-processor system is called a digital system (DS). A system consists of switches, end-nodes and communication links. The communication graph is bipartite where one part is the switches and the other part is the end-nodes. We propose two methods of a digital system diagnosis based on the fragmentation approach. The first method divides the system into 7 unified fragments, each of which contains a switch, 4 end-nodes and 8 communication links. In the second method, each of the 7 end-nodes is tested separately and every fragment includes a switch and 8 communication links. A reliable diagnosis monitor (DM) sequentially tests the components of a fragment, decodes the received syndrome and then sends the diagnosis to the maintenance staff for a recovery of the broken components. The proposed methods can be applied for diagnosis of another system with similar structure and different parameters.
|
|
