Method and apparatus for estimating the location of an unknown sound source in a structure to be monitored, wherein sounds from at least three known sound sources are detected by a plurality of detectors. At least one of peak value data and signal arrival time data derived from the outputs of the detectors are classified for each of the known sound sources and stored in a storage. Pattern differences from each of the known sound sources to a number of predetermined positions respectively, are calculated to prepare for each of the known sound sources a correspondence table indicating relationships between the calculated pattern difference and the real distances from the known sound source positions to the predetermined positions, the correspondence table being stored in a storage. The pattern differences for each of the known sound source positions are calculated on the basis of the sound signal data derived through detection of a sound from an unknown sound source and the data stored in the storage. The real distances each for the known sound source are read out from the correspondence table on the basis of the calculated pattern differences and displayed on a display device to thereby estimate the location of the unknown sound source.