Hand gesture recognition research has increased widely and rapidly to overcome various problems in human life. When in a dangerous situation humans are required to act quickly and precisely, such as being a victim of violence, accidents, disasters, and so on. In the asking process for help, it is necessary to use an appropriate safe method, communication tools, and speed in sending messages requesting help so that something unwanted does not happen. One of the ways to overcome this problem is to use sign language such as hand gestures. So that this study designs and develops a real-time hand gesture recognition system as signal for help using Deep Neural Network. The purpose of this research is to simplify, speed up, and produce a safe way that humans can use when asking for help from others without speaking when in danger happen. Base on the experiment that real-time detection and an average accuracy of 98.79%. When the system detects a hand gesture to ask for help, a request for help message will be sent to the designated emergency email. So that the benefits derived from the design and development of this system are convenience, speed, and safety in the process of asking for help when humans are in a dangerous situation. Keywords— Hand gesture recognition, Signal for help, Deep Neural Network, Dangerous Situation, Emergency Email.