Contents
I. Overview
1. What is the model-based spectral estimation
2, application scenarios, and representatives of model
II model function
Third, the spectral estimation method
1, ARMR spectral estimation
2, AR spectrum estimation
I. Overview
1. What is the model-based spectral estimation methods
The outer signal data directly in accordance with the conventional window function process 0, ignoring this data portion, hence a model-based approach to estimate the
2, application scenarios and representatives model
Application : short-term signal, the data can be estimated beyond the window, especially the noisy data
on behalf of :
- Autoregressive (AR) model (applicable peaks gluten-free, that is suitable for all-pole model )
- Moving Average (MR) model (no peaks and valleys applicable, shall apply all-zero model )
- Autoregressive Automatic Average (ARMR) model
Second, the model function
ARMR: zero pole
by AR model is introduced 2 B (z) = 1; MR model is A (z) = 1.
Third, the model solution method
1, ARMR spectrum estimation
ARMR to simultaneously estimate the parameters of AR and MR are not well established narrowband spectrum, accuracy is not high, with little in the biological signal power spectrum estimation in.
2, AR spectrum estimation
There are two ways
- Order processing (real-time entry process)
- Data processing blocks (entire waveform data processing)
Top four :( autocorrelation function does not need to directly estimate the data, with the classic method seems to be the power spectrum and autocorrelation function is the FT, i.e. Wiener - oct poured theorem periodogram method is the square of the power direct data)
Yule-Walker- produced the lowest spectral resolution, to provide maximum smoothness
Burg- can establish a stable AR model
covariance, and
Modified covariance methods
fifth order or sixth-order AR model has in some studies, the successful application of
