A Speech Privacy Protection Method Based on Sound Masking and Speech Corpus

摘要

In this paper a new speech privacy protection method based on sound masking and speech corpus is proposed. Frames in the speech corpus are selected according to secret keys and pitch of the destination speech frame, and the selected masking speech frames should have the same pitch period with the destination speech frame. Selected frames are used to mask the original speech frames and protect privacy of speech. Masked speech can be transmitted safely in communication system. By removing the effect of corresponding masking frames from the masked speech with the same corpus, original speech can be recovered. Experiment results show the proposed method gains good privacy protection effect and recovered speech quality. It is robust to speech compression by waveform coding algorithms, and helpful to systems using parametric coding algorithms for compression.