Electromyography (EMG) signals have been extensively used as a control signal in robotics, rehabilitation and health care. Creating an prosthetic hand requires many steps, beginning with a way for the prosthesis to detect the EMG signal. Unfortunately, signal detection may cause some troubles. Among them there are sensor faults that may lead to serious distortion of the signal and malfunction of the prosthetic hand. This paper is devoted to the development of algorithmic software that allows to detect faults of EMG sensors of prosthetic hand. To solve the problem modules of fault detection and signal correction were developed. These modules not only carefully supervise records from individual EMG electrodes, but, if necessary, replace them in systems if one or more signals are violated. This approach is based on signal processing methods with low computational complexity, that allows to reduce hardware requirements.