Abstract
In this paper, a method for two-sound-source tracikng is proposed based on speech spaseness. In our previous work, the two-sound-source tracking method was introduced using two successive algorithms, i.e., PAST (Projection Approximation Subspace Tracking) and IPLS (Interior Point Least Square) . In the method, appropriate initial values for the IPLS algorithm are required. In the proposed method, a weighted histogram of the DOA (Direction of Arrival) estimation results in time-frequency region is used for determining the appropriate initial values. Several experimental results in an actual environment are shown to verify the effectiveness of the proposed method.
