Environmental Sound Recognition Using Double-Level Energy Detection

Xiaoxia Zhang,Ying Li
DOI: https://doi.org/10.4236/jsip.2013.43b004
2013-01-01
Journal of Signal and Information Processing
Abstract:The performance of classic Mel-frequency cepstral coefficients (MFCC) is unsatisfactory in noisy environment with different sound sources from nature. In this paper, a classification approach of the ecological environmental sounds using the double-level energy detection (DED) was presented. The DED was used to detect the existence of the sound signals under noise conditions. In addition, MFCC features from the frames which were detected the presence of the sound signals by DED were extracted. Experimental results show that the proposed technology has better noise immunity than classic MFCC, and also outperforms time-domain energy detection (TED) and frequency-domain energy detection (FED) respectively.
What problem does this paper attempt to address?