Font Size: a A A

Design And Implementaion Of Integratied Multifunctional Voice Search Engine

Posted on:2017-03-02Degree:MasterType:Thesis
Country:ChinaCandidate:L X YuFull Text:PDF
GTID:2416330590468424Subject:Software engineering
Abstract/Summary:PDF Full Text Request
With the rapid development of communication technologies,computers and their variants have penetrated into every aspect of our lives.Communication software and applications have become an integral part of our daily social lives.These tools have improved people’s quality of life by providing convenient and fast communications;however,criminals have also used them as tools for calling and conducting criminal activities,or as platforms for criminal communicating and contacting.Processing of different kinds of voice information has become a very important part of public security work,and there are more and more voice messages need to be identified and processed.The traditional manual processing method is very laborious and time-consuming,requires extensive manpower and resources.It is impotent in handling massive amounts of data.Furthermore,deep mining and comprehensive searching of the historical voice data are very hard to perform with the traditional method,which are also prone to omissions of important information and clues.After fully understanding the actual needs of this work,an integrated voice search engine is established based on the characteristics of existing speech library.The new search engine,which integrates the cutting-edge speech recognition technology and search engine technology,effectively improves the voice data processing speed and accuracy in the public security work by converting the traditional manual identification to “smart search”.The rapid development and maturation of the speech recognition,as well as the accumulation of voice information,also provide the feasibility and experimental conditions for the multi-functional integrated voice-based search engine demanded in the public security work.The main works are as follows:(1)First,the voice processing requirements are analyzed in detail,and the idea that integrates the mainstream speech recognition technology and search engine technology to convert the manual recognition to “smart search” is proposed on the basis of the fully understanding of the actual demands.Based on the consideration of scalability and load balancing,the software architecture of the distributed multi-functional integrated voice search engine,which contains the speech analysis subsystem and search system,is proposed.(2)The mainstream technology in the field of speech recognition,such as voiceprint recognition technology,language recognition technology,gender identification technology and keyword recognition technology are integrated.The real-time analysis for the collected voice information,and the labeling of the speaker’s identification,language,gender,and keyword information included in the voice data,can not only provide immediate alarming for voices include sensitive information,but also offer the possibility for the integrated search and utilization of voice data.(3)In order to fulfill the requirements of quick searching for the voice information in the voice database,as well as the fast depth mining of historic voice data,the search engine technology is used to establish the inverted index based on the speaker’s identification,language,gender,and keyword information labeled in the voice data,as well as the time,source,and account information of the voice.(4).A high-performance multi-function voice search engine is established by integrating voice recognition and search engine technology.Jointed with the collection system,the engine can automatically analyze the language,voiceprint,and keyword information contained in the voice data file.It provides not only early warning for real-time data,but also fast and efficient queries for processed historic voice data.The full-range intelligent real-time monitoring and data mining for the voice data is achieved.In practice,good result has been achieved since the utilization of the voice search engine.It effectively solves the problems encountered in manual recognition,i.e.low efficiency and powerlessness in quick and efficient mining of massive amounts of voice data.With the voice search engine,the pressure and challenge due to the sharply increase of voice data have been alleviated,the shortage of policemen has been made up,and the strengthening of police by science and technology is achieved.
Keywords/Search Tags:Search engine, Voice mining, I-vector, PLDA
PDF Full Text Request
Related items