This article deals with the blind detection and identification of error correcting codes. More precisely, we aim at estimating the code length from a noisy bits stream in a non-cooperative context (e.g. Electronic Warfare). In this context, an eavesdropper may intercept a communication. To retrieve the sent information, one can consider identifying the channel code parameters. The method proposed here relies on a statistical analysis of a Euclidean distance matrix. It allows to estimate the code length. We address a process based on soft bits information with no a priori on the encoder used. Finally, we compare our method with a well known algorithm based on Gauss-Jordan elimination.