Application And Research On Speech Recognition Technologies In Security Monitoring System

Since the 9.11 incident, as people realize the importance of peace for the city-building, intelligent security monitoring technology has developed rapidly. In this thesis, as the current security monitoring through the use of video image analysis to achieve real-time monitoring and alarm exist deficiencies, so speech recognition technology is applied into intelligent security monitoring system for research. Through the analysis of speech information to make up for the lack of video image information, it can improve the intelligence of the intelligent security system furtherly.In this thesis, speech emotion and emotional keywords are selected as the research objects of speech recognition which play an important role in the intelligent security monitoring system. Through the analysis of the features which can represent the emotion of speech, as well as the comparison of several feature selection methods, this thesis selects the features to compose of feature vector and uses it to work for speech emotion recognition. Meanwhile, analyzing the deficiencies of the current several of common classifiers with the real-time requirement, it makes a feature contribution to inosculate near neighbor arithmetic as the way to do speech emotion recognition of this thesis.In order to recognize the emotional keywords, firstly, this thesis analyzes several representative measures which are known that there still several problems exist, then combines the needs of application environments, this article uses Biomimetic pattern recognition theory to construct the complex high dimensional covering areas to recognize emotional keywords. In the training phase, it only needs to construct the high dimensional covering areas for the feature extracted keyword samples. For the filler it doesn't need to construct any model. At recognition time, this thesis directly uses dynamic searching algorithm to search the feature extracted continuous speech without endpoint detecting and segmenting.Finally, this thesis analyzes and discusses the experiment results which can be proved it is effective. Then it makes a summary and expectation of the research.