With attention to voluntary tongue motion, which is capable of communicating the intentions of a person with a disability, we estimated the position and contact force of the tongue simultaneously using EMG signals of the underside of the jaw. We affixed a multi-channel electrode with nine electrodes to the underside of the jaw. Then, deriving many EMG signals using monopolar leads, we calculated 36 (= ₉C₂) channel EMG signals between any two of the nine electrodes. Associating these EMG signals and tongue movement using a neural network, we confirmed our ability to estimate the tongue position and contact force with precision, with a correlation coefficient greater than 0.9 and RMS error less than 10%. Furthermore, building a neural network estimating deglutition, yawning, and mouth opening, which are potential origins of false estimation, and introducing mask processing to reduce estimation error in voluntary tongue movement more than 95%, we suggest precise extraction of only the signal of that movement from EMG signals obtainable from the underside of the jaw.