Biogenic measurement has been studied as a robot's interface. We have studied the wearable sensor suit as a robot's interface. Some kinds of sensor disks are embedded the sensor suit to the wet suit-like material. The sensor suit measures a wearing person's joint, and muscular activity. In this report, we aim to establish an auto-calibration system for measuring joint torques by using EMG sensors based on neural network and sensor disks of a lattice. The Torque presumption was performed using the share neural network, which learned the data that formed the whole subject's teacher data. Additional training of the share neural network was carried out using the individual teaching data. As a result, that was able to do the neural network training in short time, high probability and high accuracy to training of initial neural network. Moreover, high-presumed accuracy was able to be acquired by this method Next, Sensor disks of a lattice was developed. EMG is measurable, checking the state of an electrode by that can measure biogenic impedance. That was able to measure EMG by sensor disks which has low impedance We measured EMG and joint torque by trial production sensor suit and torque measuring instrument. The predominancy of the torque presumption using the share neural network was check. We proposed Measurement system, which consists sensor disk of lattice. Experimental results show the proposed method is effective for the auto-calibration.