ACOUSTIC DETECTION AND CLASSIFICATION OF SOUND SOURCES USING TEMPORAL MULTIPLE ENERGY DETECTOR FEATURES

摘要

In this work, a classification method using a novel approach for acoustic feature extraction is proposed. Therefore, a multiple energy detector structure (MED) is utilized and features called temporal MED (TMED) features are introduced. The usage of an energy detector enables a pre-classification for the differentiation between impulsive and non-impulsive acoustic events. The actual classification task can be performed by using a MED. Furthermore, investigations regarding the classification accuracy using more than one microphone are presented.