Everyday, around the world crimes, like kidnapping or forced to do something to enemy’s command, are happening. General people are being the main victim in most cases. Hostage people are usually rescued by military or special force sometimes. The best way to build communication between hostage and military is by using the basic sign language of that military or special force. In this research, we analyzed 2400 images for 24 different basic signs what they use in their real mission. For this analysis, we classified their basic signs by convolutional neural network (CNN) algorithm. This research will help general people to take decision on hostage circumstances so that they can easily communicate with the military who have gone there to rescue them. We used multiple convolutional layers and get 92.50% accuracy. Using our model in any real system, a novice member of the military or special force can learn and validate his sign.