Single source bins detection‐based localisation scheme for multiple speech sources

Abstract

A single source bins (SSBs) detection based multiple source localisation scheme is proposed. This scheme is based on detecting the SSBs in mixture signals that are only derived from one source. Specifically, after proposing a 'DOA convergence' assumption, K-means clustering algorithm is used for SSBs detecting. Thus, the multiple source localisation is converted to a single source one among these SSBs. Moreover, the proposed SSBs detection is applicable to other localisation methods and not limited to specific microphone topology. Experimental results demonstrate the localisation accuracy of the proposed method outperforms the localisation approaches which are based on single source zone detection.

Multiple Speech Source Separation with Non-Sparse Components Recovery by Using Dual Similarity Determination

“…In both simulated and real rooms, the soundfield microphone was placed in the center of the room and the power of sound sources from different directions was equal in each simulation. Note that both reference methods and the proposed method were proceeded by using the localization algorithm in literature [9]. Both objective and subjective evaluation tests were adopted to assess the algorithm.…”

“…Due to the small size of the soundfield microphone [9] and the complete recording of spatial information, it is chosen as the recording array in this paper, but it should be mentioned that the proposed scheme is not specific to the chosen one. For the input mixture signals (four B-format signals of the soundfield microphone [9]), the Direction of Arrival (DOA) estimation can be obtained by a traditional localization procedure. Then, the sparse components recovery can be achieved by a clustering process of TF bins.…”

“…where n and k are the time and frequency index, respectively. µ(n, k) denotes the DOA estimate [9] at TF instant (n, k), µ i represents the estimated DOA of source i and ∆µ is a range threshold to get a better tolerance. Aiming to separate the non-sparse components by using the sparse components as a guidance, a necessary percentage of the TF components must be recovered.…”

“…However, the other TF components derived from more than one sources which called non-sparse components are ignored. Moreover, the literature [9] has verified that there are more than one sources active in many TF instants when there are multiple sources occurring simultaneously, i.e., non-sparse components occupy a major proportion.…”

Multiple Speech Source Separation Using Inter-Channel Correlation and Relaxed Sparsity

“…Further, the source separation method based on the localization procedure is not valid again. To solve this problem, some efficient localization methods have been proposed based on single-source bins or zone detection [26]. However, these methods can not eliminate the aliasing of TF components of the multiple source signals.…”