To improve the recommendation accuracy of collaborative filtering algorithms, this paper proposes a similarity measurement method based on the concept of normal cloud. First, starting from the characteristic curve of the normal cloud, combining the expected entropy curve and the outer envelope curve of the normal cloud, the paper uses the Bhattacharyya distance to characterize the similarity of probability distributions, and proposes a Bhattacharyya distance-based normal cloud similarity measurement method. Then, numerical simulation experiments are conducted to compare and analyze the performance of the proposed method with existing methods. The experimental results show that the proposed method has a better similarity differentiation capability. Finally, the method is applied to a collaborative filter recommendation system, and experiments are carried out on the MovieLens 100k movie review dataset. The results demonstrate that the proposed method is effective and feasible in collaborative filtering applications.