| In order to conform to the quick development of the voice value-added service in communication industry and satisfy people's requirements of the convenience of obtaining information, a Voice Telephone Book System based on Speech Recognition (SR) Technology is proposed in this paper. The design of the system, which is an application on service layer, is on the Voice Value-added Service integrated platform basis. And the successful design and development of the system will be a good reference for developing similar applications in the future.Firstly, this paper introduces the development and application tendency of SR, analyses the domestic and overseas researches on SR. After fully comparing the current solutions about SR appliances and considering the requirements and features of this system, a Microsoft SDK solution is presented. At the same time, this paper has done deeply research on Microsoft SDK and its Speech Application Programming Interface (SAPI) as well as its developing flow.Secondly, as anti-noise is a key technology in SR system, this paper compares some kinds of anti-noise models, which are used to enhance the SR system robust, and a proper method is presented to reduce the influences caused by background noise. After researching on current mainstream algorithmic, spectral subtraction is chosen to be used in the system.Software modularization was used in this paper and Visual C++ is the main developing tool. There are five critical modules being emphasized, that is, Communication Module, Control Module, Preprocessing Module, SR Module and Database Module, and each of their function, design idea and realization method is described in detail in this paper. In addition, the testing result is given in this paper and it shows the design of the system meets the requirements. Finally, this paper sums up the achievements and the features of the system, and pointout the further improved parts. |
PDF Full Text Request