Many amputees have maps of referred sensation from their missing hand on their residual limb (phantom maps). This skin area can serve as a target for providing amputees with tactile sensory feedback. Providing tactile feedback on the phantom map can improve the object manipulation ability, enhance embodiment of myoelectric prostheses users and help reduce phantom limb pain. The distribution of the phantom map varies with the individual. Here, we investigate a fast and accurate method for hand phantom map shape detection. We present three elementary (group testing, adaptive edge finding and support vector machines (SVM)) and two combined methods(SVM with majority-pooling and SVM with active learning) tested with different types of phantom map models and compare the classification error rates. The results show that SVM with majority-pooling has the smallest classification error rate.