Knowledge tracing is a crucial step in the field of educational data mining. While existing methods have achieved some success to varying degrees, most of them do not focus on enriching the representation of difficulty by leveraging information such as the composition of questions and the skill components in item representations, skill texts, and response correctness rates. In response to these issues, a Difficulty Representation Enriched Knowledge Tracing Model(DEKT) is proposed. The DEKT model explores the deep-level information of skill texts and combines it with question information to generate an item complexity vector. This vector, along with the response correctness rates of categorical skill texts, forms a difficulty vector that enriches the difficulty representation. Experimental evaluations using Item Response Theory models on three publicly available real online education datasets demonstrate that the proposed DEKT outperforms six classical methods in terms of AUC and ACC.