SOUND DETECTION METHOD RECOGNIZING HAZARD SITUATION

摘要

An embodiment of the present invention relates to a method for detecting specific abnormal sound sources under the environment of background noise, including the following steps of: acquiring sound sources from a microphone; separating abnormal sound sources from inputted sound sources based on NMF; extracting an MFCC parameter by the separated abnormal sound sources; calculating HMM likelihood by the separated abnormal sound sources; and determining whether the abnormal sound source is generated by comparing the likelihood of the separated abnormal sound source with a reference value. According to the embodiment of the present invention, the method can detect sounds stably under an actual environment wherein multiple noises are mixed by comparing and classifying a sound to be detected based on a non-matrix negative factorization method with a surrounding noise one-on-one.