We propose a method for estimating the speaker's head orientation in real time using multiple microphone arrays placed around the speaker. Estimation is performed using six-microphone arrays approximately 1.0 m away from the speaker. We also estimate the head direction from the observed sounds under the condition that the six microphone arrays are not synchronized, and report on the estimation accuracy. This method enables head direction estimation through remote recording, which is useful for recording interaction activities. Specifically, knowing the head direction has the advantage of being able to classify speakers participating in a conversation. The proposed method uses rough patterns of head radiation characteristics as prior knowledge and compares them with observed sound pressure levels obtained through distributed microphone arrays. Since it is a comparison method using frame energy, it is a method that can robustly estimate direction even when microphone arrays are not synchronized. The estimation accuracy of direction estimation classified into eight directions of 0, 45, 90, 135, 180, 225, and 315° was approximately 70%.
Takahashi et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: