Overlapping Acoustic Event Detection via Perceptually Inspired the Holistic-based Representation Method

ABSTRACT

....By using Mel-scale in a dictionary learning, we show that the information carried by low frequency components more than high frequency components and dealt with a low complexity. Also, the proposed holistic-based representation method avoids the permutation problem between another acoustic events. Based on these benefits, we confirm that the proposed method of Mel-scale with NK-SVD delivered significantly better results than the conventional methods.

2. Proposed Method for Overlapping Acoustic Event Detection

.... As observed earlier, an auditory system recognizes acoustic events non-linearly in frequency domain. Therefore, we propose a nonnegative dictionary learning method based on Mel-scale which extracts relevant features nonlinearly in frequency domain. Thus, higher resolution is applied in lower frequency regions that contain more salient frequency changes. It was based on the known evidence that the information carried by low frequency components of the speech signal is phonetically more important than carried by high frequency components. However, unlike a speech signal, a high frequency signal could cause the performance deterioration.

Conclusions

...Thus, higher resolution is applied in lower frequency regions that contain more salient frequency changes. However, unlike a speech signal, a high frequency signal could cause the performance deterioration.

...

Also, the proposed holistic-based representation method avoids the permutation problem between another acoustic events.