This paper introduces our system submitted to Voice Privacy Challenge 2022. We adopted the following methods to improve the traditional methods. Firstly, the adversarial anonymization method was used, further hiding speaker information. Then, we extracted the embedding from the encoder of the transformer based ASR systems because ASR has rich speaker information, so we do not have to train an individual speaker recognition/verification system for speaker embedding extraction. Experimental results prove that the proposed methods can be used for speaker anonymization tasks.