The current paper proposed and investigated the head motion recognition idea based on four capacitive sensors and deep learning models. The proposed system was designed to empower a tetraplegic person to control a remote device or an intelligent wheelchair. The capacitive sensors were placed around the neck using a necktie, which each volunteer who participated in this experiment was easy to use. The results show that the best-proposed deep learning model can determine each activity with a classification rate equal to 89.29% using capacitive raw data. During the experiments the deep learning models provided accuracy values in the range of 56.25% to 89.29%.