An error correction encoding method is provided for encoding in parallel source digital data, having the form of a frame, wherein said data can be classified into N classes, where N is an integer at least equal to 2.The encoding method includes: a first recursive systematic convolutional encoding step of data to be encoded, formed by the data of the class 1; and an implementation of the following steps, for each n ranging from 1 to M, where M is a positive integer equal to or lower than N−1: nth mixing of a set formed by the data of the class n+1 and the systematic data of the preceding encoding; and (n+1)th recursive systematic convolutional encoding of data to be encoded, formed by the result of the nth mixing. Also disclosed is a related decoding method, as well as an associated encoding and decoding devices.