Turbo convolutional encoders are one of the newest types of channel encoders. A puncturing block is used to create the possibility of encoding at different rates by an encoder. The successful implementation of blind parameter recognition process of these encoders for extracting information from the received bit sequence is essential. Since the structure of this encoder is generally created by combining two convolutional encoders and a permutation, separating the output data of convolutional encoders in normal and punctured conditions can make it much easier to recognition the structure of the encoder. In this article, using the technique of partitioning orthogonal vector matrices, a method for classifying members of the turbo convolutional encoder code word into convolutional encoder outputs 1 and 2 has been presented. By running the proposed algorithm on the output of several different simulated encoders, it has been shown that this method is also executable in punctured conditions.