A novel background noise estimation in adverse environments

This paper addresses the problem of noise estimation in adverse environments. It contains two new parameters, the mel-scale frequency bank entropy (MS-entropy) and the delta minimum frequency energy (delta-MiFre). The MS-entropy parameter can calculate the uncertainty of spectral magnitude, and the delta-MiFre can process the variation of temporal spectrum energy. We use these two parameters to avoid the wrong background noise estimation. Hence the background noise level estimated by the proposed method is more precise than that estimated by the pure spectrum energy method.