This technology is used to determine emergency situation by recognizing sound that can be generated in various emergency situations, such as the sounds of screaming, calling for rescue and whistling for self-protection. It is used for prevention of and real-time response to crimes in certain areas where they have a higher tendency to occur, such as parks, as well as at home and school.
The stochastic model recognition prediction algorithm based on the analysis of natural frequency pattern of voice and sound builds a state machine and a decision-making tree for uncertain emergency situations by analyzing the correlation between crime situation factors, modeling entropy filtering volume and knowledge-based emergency type information and using a correlation analysis engine.
- Noise robust Feature extractor
- Eigen-vector-based characteristic parameters
- Patter Classifier
- GMM/HMM/SVM-based probability statistics modeling technology
- Powerful classification algorithm using a bootstrap technique
- Decision rules based on reliability
- Emergency decision tree update rules
- Diverse types of voice/ sound DB packages
- Sound DB for calling for rescue/ screaming/ emergency situations
- Sound DB for whistling/ sirens/ vehicle horns