This　paper　proposes　a　multiple　source　localization　method　by　using　the　energy　weighted　＇single　source＇　zone　detection.　Specifically,　based　on　the　well-known　theory　that　the　sum　of　two　values　will　be　more　close　to　the　bigger　one　if　the　two　values　vary　widely,　we　suppose　that　the　time-frequency　(TF)　component　with　high　energy　of　a　source　is　not　easy　to　overlap　with　the　one　of　other　sources.　Therefore,　after　detecting　a　＇single　source＇　zone　(SSZ)　with　a　slightly　lower　threshold,　we　can　find　more　SSZs　in　a　short　period　for　localization　and　improve　the　low-　delay　property.　The　pseudo　＇single　source＇　zone　problem　caused　by　the　lower　threshold　is　solved　by　weighting　each　DOA　estimate　based　on　its　TF-bin-energy　in　the　zone.　The　proposed　method　further　eliminates　the　crosstalk　(cross-talk　distortion　results　when　an　additional　TF　component　from　another　source　is　incorrectly　assigned　to　the　separated　source.)　among　multiple　sources　and　the　evaluations　reveal　that　the　proposed　method　achieves　a　higher　localization　accuracy　compared　with　the　existing　＇single　source＇　zone　detection　based　methods.　©　2017　IEEE.