Direction-of-arrival (DOA) estimation based on microphone arrays has been a hot research topic in recent years. Transfer function (TF) based DOA method performs well because it considers both time difference and intensity difference. However, obtaining transfer function is a difficult task and transfer function based method is susceptible to noise. In this paper, an autoencoder network structure is proposed for DOA estimation task. The network is used to learn the characteristics of the transfer function, which considers both time difference information and intensity difference information for DOA estimation. The proposed unsupervised training method helps minimize the burden for labeling training data. The evaluation experiments show that our method performs better than TF-based method in the noisy environment.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
