The purpose of this paper is to develop a method for automatically detecting vocalizations of laying hens from a large amount of original data. First, the original sound data were cut into 1 sec of length clips by using sound analysis software. Then, 130 clips including calls of Hy-Line Brown laying hens and 132 fan noise fragments were selected. Finally, an algorithm was developed for automatically identifying vocalization of laying hens and fan noise based on the discrepancies of 2 kinds of sounds’ power spectral density in 1000~1500Hz. The results showed that a 95% overall correct classification ratio was achieved and 93.3% of the vocalizations of laying hens were correctly identified. This approach can improve the efficiency during sound analysis and reduce the storage and transmission of useless sound.