The present article is a series of publications dedicated to the research of smart fabric sensors integrated into socks and is also part of the project aimed at developing the measuring system based on smart fabric supplied with sensors and intellectual data processing. The aim of the article is to perform a practical study on the application of Self-Organizing Map to smart textile signal clustering. Within the framework of the research, different approaches to the organization of network training are explored. A method for encoding an input pattern is also proposed. It has been established that the network is able to recognize the signal as a good step, a bad step, and an unrecognized step. The primary classification allows further selecting specific algorithms for a detailed analysis of good steps and bad steps. The detailed analysis of bad steps is the key to solving the problem of revealing of an athlete’s special type of fatigue, leading to injuries.