Towards Smart Citizen Security Based on Speech Recognition

Citizen security is a fundamental determinant of the well-being of households and communities. Robberies, rapes, domestic violence, gender violence, kidnapping, among others, are examples of situations of insecurity having a negative effect on citizens and governments worldwide. This paper deals with providing a surveillance technology based on speech recognition for citizen security without affecting the privacy of citizens. To this aim, a prototype software component for the automatic detection of situations of insecurity based on speech recognition, natural language processing, and pattern recognition is proposed. This component provides an automatic surveillance service that allows detecting in real time the occurrence of situations of insecurity, generate security alerts to help mitigating such situations, and report/record the occurred events. The software component minimizes the invasion of privacy of citizens, since speech recognition is performed entirely local, with no access to the Cloud. It also avoids the registration of any dialogue or conversation in which no situation of insecurity has been detected. The implemented prototype is evaluated with simulated situations of insecurity taken from Argentinian's movies and with real cases taken from videos recording criminal acts in different cities of Argentina. Our preliminary results show the feasibility of the approach.