The Task of Recognition of Violent Situations Using Automatic Systems and Methods of Artificial Intelligence

The article discusses the task of development of AI computer software with the purpose of automatizing the stage of indication of violence markers that provides an operator with processed information and helps him to analyze the situation and make a decision. The advantage of such systems is dramatic widening of the scope perceived by an operator since the system draws his attention only to the points of emerging or possible breaches of public order. An operator is provided with the opportunity to monitor the situation in numerous observed locations in the real time and to take necessary measures for prevention of a threat escalation or eliminating consequences of an accident. The current trend in development of such automatic systems is the transition from visual information processing to multimodal analysis based on combined audio- and video- streams translated from the scene of action. It is shown that simultaneous processing should begin at the first stages of the analysis since it is most rational not to summarize data from independent processing systems but to “merge” streams of audial and visual information and process them together as a single stream of data. Thus in modern developments of behavior recognition systems the model, close to psychological concepts of human perception, is implemented.

1. Aniwenko S., Shaposhnikov D., Podladchikova L., Kamli R., Suholencev K., Gao K. Monitoring dvizhenij golovy s pomow'ju foveal'nogo podhoda i detektirovanija lokal'nyh licevyh opornyh tochek // http://nisms.krinc.ru/papers/PRIA_9_rus.pdf
2. Gafurov A. O. Algoritmy ocifrovki zvuka i nejrosetevye metody raspoznavanija slov i jemocij cheloveka ili zhivogo suwestva v intellektual'noj nejroinformacionnoj sisteme «NejroKiber» // http://infgeoservice.narod.ru/publik2.html
3. Lopatina A. D. Vydelenie oblasti lica s pomow'ju kombinacii metodov cvetovoj i jarkostnoj segmentacii // Vestnik UGATU. Upravlenie, VT i I. 2009. T. 13. № 2 (35) // http://www.ugatu.ac.ru/publish/vu/stat/ugatu-2009-2(35)/24.pdf
4. Poljakova M. V., Iwenko A. V., Hudajberdin Je. I. Porogovo-prostranstvennaja segmentacija cvetnyh teksturirovannyh izobrazhenij na osnove metoda JSEG // AAJeKS. 2010 №1(25) //http://aaecs.org/polyakova-mv-ishenko-av-hudaiberdin-ei-porogovo-prostranstvennaya-segmentaciya-cvetnih-teksturirovannih-izobrajenii-na-osnove-metoda-jseg.html
5. Sajt «Rechevye Tehnologii» // http://speetech.by/press/analytics/1
6. Datcu D. Multimodal Recognition of Emotions // Wőhrmann Print Service, 2009.
7. Fasel B., Monay F. & Gatica-Perez D. Latent Semantic Analysis of Facial Action Codes for Automatic Facial Expression Recognition // http://www.idiap.ch/~gatica/publications/FaselMonayGatica-acmmm-mir04.pdf
8. http://affect.media.mit.edu/
9. http://www.face-and-emotion.com/dataface/facs/description.jsp
10. http://www.face-rec.org/databases/
11. http://www.ti-eng.ru/technology/imagerecognition/
12. Yang Zh. Multi-Modal Aggression Detection in Trains. Delft: TU Delft Mediamatica, 2009.