Method and device for rapid identification of explosion source information based on acoustic characteristics

Abstract

This invention discloses a rapid identification method for explosion location information based on acoustic features. The method includes: S1: Measuring the parameters of the propagation medium at the scene; S2: Collecting environmental noise; S3: Collecting explosion sound signals corresponding to different explosion sources to form an original sound signal database I; S4: Preprocessing the signals in database I to extract short-time energy features, cepstral coefficient features, explosion spectral texture HOG features, and explosion spectral color moment features, and establishing a training feature database D; S5: Training an information vector machine model using the four types of fusion features in the training database D; S6: Measuring the explosion sound signal at the scene and preprocessing the signal; S7: Extracting the four types of fusion features from the preprocessed signal at the scene and inputting them into the trained IVM model to determine the explosion location information; S8: Dynamically updating the original training database D with the accurate prediction samples from the IVM prediction results as new training samples, and retraining the IVM.

Images