If the matrix is the empirical covariance of square sum signal u, then u is called n-order persistent excitationsignalThe continuous excitation condition is added to the input signal u so that the finite pulse can be estimated by the least square method
Picture 1
Impulse response model
Picture 2[1]
Excitation signal
Announce
edit
The excitation signal isInput signal, mainly to reduce the square wave frequency and stack the voltage.
In the square wave polarography, the square wave voltage is continuous, while in the pulse polarography, when each drop of mercury increases to a certain time (such as 3 seconds), a 10-100mV pulse voltage is superimposed on the DC linear scanning voltage, and the pulse duration is 4-80ms (such as 60ms);
Relation to frequency
Announce
edit
U (t) isnOrder PE, then the frequency spectrum in the interval [- pi,+pi] has at leastnThe point is not zero
twoOnenLevel PE signal cannot ben -The 1st order moving average filter can not be filtered to zeroNote: This is also used as the definition of PE in some places
Judgment method
Announce
edit
The conditions for determining whether a condition is a continuous incentive are as follows:
Judgment conditions
application
Announce
edit
The relationship between continuous motivation and the performance of the deterministic learning algorithm:
By adoptingRadial basis functionNeural network, a set ofNonlinear system identificationAnd the deterministic learning theory of dynamic pattern recognition is proposed.Based on the deterministic learning theory, the following conclusions are drawn:
(i) The learning speed is improved with the increase of continuous incentive level;
(ii) There is an optimal learning speed;
(iii) The learning accuracy is also improved with the increase of continuous incentive. In particular, when the level of continuous incentive is large enough, the local accurate learning effect will reach the level of required accuracy. However, when the level of continuous incentive is low, the learning performance will deteriorate.[2]