The distribution of hexa-structures in secondary structure sequences of different classes of proteins has been found. Based on this, two methods for the recognition of the structural class of a protein are proposed. The first is the method of Mahalanobis distance which is based on the frequencies of tri-structures in secondary structure sequence. The second is the method of diversity measure which is based on the frequencies of hexa-structures that are regarded as the source of diversity. The prediction has been done in a set of 1 130 proteins of four classes, namely α-class, β-class, α/β-class and multi-domain protein. The successful rates for two recognitions are about 81% and 83% respectively. The method introduced here also gives an approach to predict the compact structural domain of proteins.