Tongue-machine interface (TMI) is a feasible way between the assistive technologies and paralyzed individuals who have lost their abilities to communicate with the environment. Researchers have presented equipment based tongue-machine interfaces to reach a reliable and speedy system. However, this kind of interfaces may occur a way of obtrusive, unattractive and unhygienic for disabled persons. In this research, we intended to propose a natural, unobtrusive and robust glossokinetic potential signals (GKP) based TMI exploring the success of the novel machine learning algorithms. The tongue is bound up with cranial nerves to the brain, which can escape from the spinal cord injuries in general. Moreover, the tongue has highly capable of sophisticated manipulation tasks with less perceived exertion in the oral cavity and gives degrees of privacy. In this study, ten naive healthy subjects have attended who were between 22-34 ages. Decision Tree (DT) and k-Nearest Neighbors (kNN) algorithms were used with Mean-Absolute Value (MAV) and Power Spectral Density (PSD) methods. Moreover, Discrete Wavelet Transform (DWT) was implemented to reveal the theta and delta subbands. In the study, the highest value was provided as 96.77% by the k-Nearest Neighbor algorithm for the best participant. Furthermore, the GKP-based TMI may be an alternative system for the limitations of the brain-computer interfaces. It is well-known that EEG deficits are major concerns for brain-computer interfaces.