Fig. 5
From: Multichannel speaker interference reduction using frequency domain adaptive filtering
MAEC approach in the meeting scenario for speaker m = 1. Thereby, the box 
denotes again the buffering of the previous K − R samples, while 
shows the elimination of K − R samples w.r.t. the OLS constraint. Additionally, 
depicts the replacement of K − R samples by zeros. The RIRs H1,2(ℓ,k) and H1,3(ℓ,k) of interferer signals \(S^{\prime }_{2}(\ell,k)\) and \(S^{\prime }_{3}(\ell,k)\) are estimated (\(\hat {H}_{1,2}(\ell,k), \hat {H}_{1,3}(\ell,k)\)). Afterwards, microphone signals Y2(ℓ,k) and Y3(ℓ,k) are multiplied with \(\hat {H}_{1,2}(\ell,k)\) and \(\hat {H}_{1,3}(\ell,k)\), respectively, and are then subtracted from the microphone signal Y1(ℓ,k), resulting in \(\hat {S}_{1}(\ell,k)\). By means of the error signal E1(ℓ,k), the estimation of the RIRs is adapted. To obtain \(\hat {S}_{2}(\ell,k)\) and \(\hat {S}_{3}(\ell,k)\), the inputs are changed accordingly