Spectral analysis of nonstationary signals
| When |
Jan 31, 2022
from 11:00 to 11:00 |
|---|---|
| Where | Online seminar |
| Attendees |
Adrien Meynard |
| Add event to calendar |
vCal
iCal
|
In signal processing, spectral analysis aims to decompose a signal as a superposition of periodic signals. This takes on its full meaning for stationary signals, i.e. possessing invariance properties by time translation. One thinks, for example, of notes from musical instruments (the pitch of the sound being constant), or the sound of a motor rotating at constant speed. We have developed an approach for the spectral analysis of nonstationary sounds, such as an accelerating engine noise, or a melody sung by a human voice. The approach is based on modeling the signal in the form of a stationary signal whose stationarity is broken by a transformation (for example an acceleration of the speed of an engine). Spectral analysis then amounts to estimating both the transformation and the spectral characteristics of the underlying stationary sound.