The main activities of the laboratory are the development of algorithms, methods and software systems for the solution of the following problems:
1) Analysis and modeling of information processes;
2) Digital signal processing;
3) Automatic synthesis of the speech by text;
4) Automatic speech recognition;
5) Automatic processing of text data and extraction of knowledge from texts;
6) Realization of computer vision systems.
The systems of synthesis and speech recognition, speech database, tools for automatic generation and marking the speech database, automatic processing of texts in Kazakh language have been implemented.
The obtained results:
- A method of phonetic-acoustic classification of the speech signal, which allows to unify the acoustic-phonetic structure of the language, has been proposed.
- A specialized language of phonetic representation which allows to set and describe phonetic and intonational forms of the speech has been created. Input data of the model are described with its help. That allows flexible intersystem interaction.
- An algorithm of phonetic transcription of texts in Kazakh has been designed.
- A framework, implementing the unified methods of speech signal synthesis by a phonemic text regarding Kazakh language, has been developed.
- Tools in the form of an information system for automatic generation of digital voice portrait of a person (DVPP) based on the digital recording of a natural speech signal pronounced by a person whose digital portrait is being formed; the DVPP standard defining the structure of the used models and their parameters, acceptable ranges of these parameters’ values changes, and also presentation format, storage and data exchange have been developed. Standard procedures of DVPP use in speech synthesis systems, recognition and voice identification have been defined in this standard.
- Algorithms to solve the problem of computer speech synthesis based on DVPP technology, and program implementation of the corresponding information system have been built.
