The kurtosis of spectral envelope peak as auditory cuing of sound identification

摘要

Spectral envelope has been identified as an important aspect of timbre characteristic of sound sources. Sound from different sources, such as voice or traffic noise, usually has different timber characteristics, which makes a kind of sound be distinguished from the others in a competitive sound environment by the auditory system. How to transform such an abstract feature into exercisable variables is still unknown. In this research, a concept of “kurtosis of spectral envelope peak (KSEP)” was proposed as a characteristic parameter to identify the specific sound from different sources. By analyzing the signals from different kind of sound sources, we found that KSEP can distinguish sound sources well, even when the sound signals are in the same frequency band. Based on the evidence, KSEP was further used as criteria to distinguish speech from noise to reduce noise form noisy speech, and the SNR of the noisy speech was effectively improved.