My observation is that for testing the performance of LMS algorithms always Gaussian white noise with zero mean is used as input signal (for both, the unknown and the adaptive system). Sometimes a cosine function of a certain frequency is added. In practice, however, the input signal is given by the application. The power density at f=0 could be much higher than at other frequencies. Standard adaptation algorithms seem to have difficulties to find the correct filter coefficients in such cases. Is there any special adaptation algorithm know for this problem?
Dear Tilo,
I suggest you to see links and attached files in topics.
-Adaptive LMS Algorithm System Identification using LabVIEW (PDF ...
https://www.researchgate.net/.../267798180_Adaptive_LMS_Algorithm_System_Iden...
- LMS Adaptive Filters - MATLAB & Simulink - MathWorks
https://www.mathworks.com › ... › Adaptive Filters
- Noise Estimation in Electroencephalogram Signal by Using Volterra ...
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4528358
- Reference Data for Engineers: Radio, Electronics, Computer, and ...
https://books.google.dz/books?isbn=0750672919
Best regards
Hi Tilo,
I wonder if the problem is only with the estimator... what if you include a constant (e.g. one) regressor in the filter to cope with the large dc value. Such a value tends to move the problem away from linearity (in terms of the superposition property) and some models/algorithms may lose performance.
Dear all,
thank you for your responses. Unfortunately, the answers do not lead in a feasible direction. The linked articles are either very basic or use Gaussian noise as input signal for the simulation (what I do not want). It is also not clear to me, how GA (genetic algorithm?) could help in system identification. I have tried to extend the model function (the FIR filter) by a bias parameter as suggested by Luis Antonio Aguirre, however, there is obviously no force driving this parameter to the mean value of the input signal. Besides, an additional parameter would not correspond to the impulse response of the unknown system.
Is there anybody with other suggestions?
Many thanks!
- Badges
- Science topic
- Similar topics
- Correction