Abstract
This paper presents a method for silence deletion performed on audio signals extracted from digital video before and after MPEG compression in order to get the relevant part of the sound stream and avoid silent passages at the edges.
In fact, the MPEG compression damages the sound stream by adding an audio silence in the beginning and at the end in some cases such as when using MPEG layer III for the audio coding.
So, we propose in this work an approach of silence deletion allowing us to remove the low dynamic passages from the beginning and the ending of the audio signal.
Consequently, we get rid of the silent intervals added by MPEG compression and the very frequent soundless passages at the edges of audio signals.
The phase of silence deletion is performed in two steps. First, a pre-processing step which consists on searching optimum thresholds is applied. Second, a step of silent samples deletion is performed according to the defined threshold.
In order to evaluate our approach, we propose to calculate the similarity between the two cut signals: the original and the compressed/decompressed one using the signal to noise ratio metric (SNR). The SNR metric should be as high as possible, meaning that no shifting exists between the two signals.