Audio-based surveillance systems can be used in public places to detect abnormal events because such events are usually accompanied by abnormal sounds, such as screaming, explosions, gunfire, and crashing sounds. Audio-surveillance systems can supplement video surveillance. This paper proves that a T-distribution model is highly suitable for describing a wide range of typical background noise distributions encountered in public places. Background noise in public places can affect feature extraction for abnormal sounds when Local Mean Decomposition (LMD) is used as a signal-processing tool. The authors first confirm that the background noise obeys a T-distribution using Kolmogorov-Smirnov hypothesis testing. The authors propose an improved LMD method based on the T-distribution to enhance features extraction. They add particular production function components of inhomogeneous random noise obeying a T-distribution to the abnormal sound in a nested manner and then take the ensemble means of the obtained production functions as the decomposition results. This alleviates the mode mixing inherent in LMD. Additionally, the algorithm replaces moving average operation with a linear spline to reduce the iteration required in LMD from triple-loop iteration to double-loop iteration. Experimental results demonstrate that the proposed method outperforms commonly used methods in terms of the classification rate and computational cost.
Click to purchase paper as a non-member or you can login as an AES member to see more options.
No AES members have commented on this paper yet.
To be notified of new comments on this paper you can
subscribe to this RSS feed.
Forum users should login to see additional options.
If you are not yet an AES member and have something important to say about this paper then we urge you to join the AES today and make your voice heard. You can join online today by clicking here.
