Own voice detection with near field head related transfer function based on frequency domain binaural model
Taira Onoguchi, Yoshifumi Chisaki, Tsuyoshi Usagawa · The Journal of the Acoustical Society of America · 2012
On the binaural hearing assistance systems with directivity control, it is difficult to distinguish user's own voice from the target speech located in front of the user. Therefore, emphasized user's own voice degrades the quality of hearing assistance. Because the own voice arrives from the specific angle in the median plane in the near field, it is difficult to distinguish whether the incoming sound is own voice or not. However, in the previous works, the elevation angle of signal source can be estimated even on median plane by means of artificial neural network based on the interaural level and phase differences. In this paper, a new method to detect own voice is proposed. This method utilizes the artificial neural network to discriminate own voice from any of signal sources located far than 1.5m from the user. The performance of this system is examined by simulation.