Secret key generation scheme that utilizes the randomness of wireless channels is a very promising alternative to establish a secure communication path. Some wireless devices are used to obtain the channel parameters. But the problem that occurs is imprecise measurement results, causing a secret key difference between the two users. To overcome this problem, preprocessing method is used before the data is quantized, to increase the similarity (reciprocity) channel parameter measurement results. In this paper, the effect of the use of one pre-process method, which was hierarchical clustering, on the performance of secret key generation scheme in an indoor environment with two variations of scenario was investigated. The results show that the use of the pre-processing method can improve the similarity parameter of the channel measurement results as indicated by the increasing of correlation values up to 20%. In addition, the resulting bit mismatch also decreases with the value of key disagreement rate (KDR) maximum of 20%.