Speaker:
Mark Kahrs
- Helsinki University of Technology
Thursday, March 15, 2001
10.00
The Teager Energy Operator (TEO) is a nonlinear time domain operator.
This paper reviews the principles of the TEO along with specific
audio applications.
We first review the TEO as well as the Discrete Energy Separation
Algorithm (DESA). We demonstrate how the TEO can be used to
detect missing samples or other kinds of transients.
We then compare the well known Phase Vocoder with Teager Energy Operator
using synthetic signals. The TEO performs better under low noise
conditions and when using a Gabor bandpass filter. We also
compared the Hilbert Transform Instantaneous Frequency detection
with the TEO and again the TEO shows better performance in the
detection of AM and FM signals.
We are interested in two aspects of musical instrument modeling: 1)
separation of the excitation and resonator and 2) modeling of performance
parameters such as tremolo (AM) or vibrato (FM). We show how ``Energy
Separation'' can be used to determine synthesis parameters for an
excitation/filter model, and also be used to follow tremolo and vibrato.
The use of the Teager Energy Operator shows great promise for
analyzing audio signals with performance nuances.
