Research On Identification Of Dangerous Situation Based On Speech Emotion

With the rapid development of information technology and people for security consciousness unceasing enhancement, how to protect people’s life and property safety become the important subject in the new period. Now in order to strengthen security and municipal administration in some crowded public places, such as railway station, airport, square cameras installed more and more. According to the report, there are about 4.2 million cameras in public places and the number is increasing. In our country, with the improvement of living standards and the security awareness higher and higher, otherwise, "safe city" and other large urban informatization project construction, the application of the video monitoring is also growing sharply.Video monitoring system has two limitations:first of all,it depends on artificial analysis and decision heavily. And the number of people can effective surveillance cameras and time is limited, so the stability and reliability of the monitoring system has a problem. Secondly, it’s unable to realize real-time monitoring alarm, because ordinary video monitoring cannot be done automatically screening and processing of data, many cases is used for collecting evidence afterwards,not achieve real-time alarm. Now with the development of speech recognition system and the popularity of the application of the security alarm, how can the speech recognition or the study of speech better application in the field of security, at the same time make up for the inadequacy of video monitoring system, to make security work achieve better effect, is the aim of this paper.In this paper, starting from the method of speech recognition and artificial intelligence, based on multiple voice database,from a new angle of view sound properties is the study of the characteristics of voice dangerous situations, and applied to security alarm system, to improve people’s life and property safety. Main contributions of this article has the following several aspects:first, only put people on the first place, from a new perspective to study voice sound properties. Second, combining speech recognition and speech emotion angle together,to study the characteristics of sound dangerous situations, and apply it to security alarm system.Third, through the establishment of the artificial neural network model, combining multiple stage after test of emotional speech databases and abnormal sound databases, and through the acquisition of the film and television dangerous conditions of voice to make up the voice of the database, then pick up the extracted essence of speech feature parameters, one or more characteristics composed of vector group, proceed experiment,in compare with people often adopt MFCC parameters for experiment result.