| Abstract [eng] |
It is well known that speech recognition-based interfaces could have great value in many applications. In recent years, speech recognition technologies have been widely applied throughout information technologies. Therefore, the application of speech recognition in information technologies is an extensively explored area. Systems for the recognition of the spoken language are applied in various areas. The goal of this thesis is to create a hybrid recognition technology for Lithuanian voice commands that connects two or more speech recognizers. It is expected that, as a result of connecting different recognizers, if one recognizer makes a mistake then another/others will make the correct decision. The results of this research could be applied and used in the development of ASR systems for applications involving the recognition of disease names according to their codes (using ICD-10-CM), consisting of one letter and several digits. Examples of codes containing only digits that could be recognized include PIN, personal identification codes, etc. Another potential application is the recognition of codes containing only digits through the telephone. |
