An Audio Event Detection Approach Based On Cross-Correlation In Selected Frequency Bands of Spectrogram

Audio event detection (AED) systems have various applications in modern world. Its applications include security systems, urban management and automatic monitoring in smart cities, online multimedia processing in virtual space, natural language processing, etc. The noise and background sound of environments vary greatly, which is why designing an AED method requires an event/environment based approach. In this work, a spectrogram-based approach is proposed that uses the spectral characteristics of audio signals and cross-correlations to build a dictionary of effective spectrogram frequency bands and their patterns in different audio events. The proposed approach removes background sound from the input signal through a preliminary pre-processing step. In the next step, a mathematical-statistical analysis is used to specify the effective frequency bands of the spectrogram in each audio event. The specified frequency bands and their patterns are saved in dictionary. The proposed approach was evaluated in DCASE 2013 database, and its efficiency was compared with the state-of-the-art research in the field. According to the results, about 30% of the frequency bands in the spectrum are useless, and only 70% of the frequency bands can be used for an AED system. The pattern of selected frequency bands varies due to the type of event, which can effectively help reduce errors and increase the accuracy of different AED methods. The proposed approach can be used as a suitable feature selection approach in all methods which use spectral features.