The invention provides a muscle movement unit extraction method based on a neural network; a muscle movement unit is extracted through an array type surface electromyographic signal by adopting the neural network in a muscle dynamic contraction process. Muscle movement unit waveforms are classified by an ART2 neural network, neural network training is not needed in the classification process, actual waveform signals are directly input for recognition, the recognition process is simplified, and the operation time is saved. In the classification process, for replying to changes of the movement unit during muscle dynamic contraction, the classification template is dynamically updated. When the muscle movement unit is confirmed, the waveform propagation characteristics are considered, and theaccuracy of extracting the muscle movement unit is improved. The method is easy to implement and meets the requirements of practical application.