Suspicious Speech Tracking Using Machine Learning (SSTRUM)

Abstract

Terror activities are a bitter truth of our society. It has always been our governments first priority to avoid such activities by secretly spying the terrorists by different means. In the past decade, since ways of communication were not that strong so they were tracked manually. Now, with increasing advancement and automations, an increase in terror activities has been spotted, especially on special events where a large community is gathered. Definitely telephonic communication plays a major role in it. Security agencies are manually tracking every call to check if the speakers are trying to plan out any terror activities so that it could be avoided. However, this requires a large human resource since on average a specialized intelligence person could efficiently listen to telephonic conversations for 10 hours only. Government is now forced to even cut off telephonic signals on special gatherings. However, the problems rising due to technology should be dealt with technology. SSTRUM – Suspicious Speech Tracking Using Machine Learning is our prototype solution to such technology related telephonic problems. This system will be capable of automating the process of speech tracking for Pashto language to avoid ill activities. With the use of machine learning algorithms our system will have accuracy of over 85% in the tracking of pre fed suspicious words, commonly used by trouble mongers. A standalone device has been created for this purpose, which will be able to run our system at any place allowing the mobility, also serving any special occurrence efficiently. Our desktop application will be compatible with multiple operating systems with a user friendly interface which will help user track their audios with rapid response time.