The purpose of this study is to solve the bottleneck problem of discriminative object tracker using correlation filter. On the period of learning and updating on correlation filter, errors are likely to be induced into filter, and fatal errors will finally cause tracker inefficiency. Using bidirectional clockwork recurrent neural network to construct confidence map to identify whether the moving target is blocked, a new discriminant tracking algorithm is proposed by integrating clockwork recurrent neural network and kernel correlation filtering. The proposed CKT algorithm uses confidence map to guide the state updating of the clockwork recurrent neural network and optimize the learning process of subsequent kernel related filters. The above measures assist in solving the self-learning problem of the correlation filter in the learning process. Compared with the mainstream object tracking methods on standard testing videos involving VOT2016, the tracking experiments demonstrate that the CKT algorithm respectively lists first, first, fourth and fifth rank on the tracking data A-rank, EAO, R-rank and EFO. On the complicated scenes such as object occluded, object speed drastically changing and light change, etc., the CKT algorithm has better tracking performance than the GGTV2, STAPLEp, CCOT, sKCF and SSKCF algorithms. The proposed CKT algorithm is especially suited to machine learning process where samples are continually acquired and memory storage is limited.