Sensor localization is one of the big problems when building large scale indoor sensor networks. We are developing ZigLoc, a sensor localization system using Wi-Fi (IEEE802.11) APs (access points) as references , . ZigLoc measures RSS (received signal strength) of Wi-Fi AP signals to localize ZigBee (IEEE802.15.4) sensor nodes. However, ZigLoc exhibits low accuracy because of inaccurate RSS measured on a single narrow-band ZigBee channel. In this paper, we present a highly accurate sensor localization system MultiZigLoc, which is an extended system of ZigLoc. Our key idea is to employ RSS measured in multiple ZigBee channels in fingerprinting localization. The RSS in multiple channels is dependent on the measured location because ZigBee uses narrow-band channels. Narrow-band ZigBee communication is highly affected by frequency selective fading, whose influence is dependent on channels and locations of both Wi-Fi APs and a sensor. We utilize fingerprints that separately handles RSS in multiple ZigBee channels to employ channel specific features. We conducted initial evaluations using RSS measured in a practical environment. The evaluations reveal that MultiZigLoc improved the localization accuracy by more than 10 points.