In this paper, a new class of gradient-based algorithm by using variable step-size technique for a second-order adaptive IIR notch filter is presented. An adaptation stepsize parameter for the algorithm is worked by the output signal and the gradient signal. The adaptive algorithm is used to detection of a sinusoid with additive white Gaussian noise, impulse noise and cancel 50-Hz interference in the recording of electrocardiograms signal. The performance of this algorithm proved here that gives high convergence speed, high impulse noise robustness and good efficiency of cancellation 50-Hz. Finally, the results of computer simulation are given to demonstrate the performance of the proposed algorithm for adaptive IIR notch filter.parameters from the derivation of the transfer function. This algorithm is used to estimate frequency of the sinusoidal embedded in impulse noise and cancel 50-Hz interference in the recording of electrocardiograms (ECG) signal. The simulation results for several situations are tested by using the computer programs to show that the proposed algorithm is an attractive adaptive algorithm.
A NEW VARIABLE STEP-SIZE ALGORITHMIn this section, we first describe the QLMp algorithm that was proposed in ref.[3] and then the proposed algorithm. To simplify, assuming that the input signal to the ANF is a single sinusoidal corrupted by Gaussian white noise, i.e., 1. INTRODUCTION . x(n) = Asin(won + 4) + v(n),n = 0,1,2 ,... (1) Digital notch filters have many applications such as the detection of sinusiods in noise or eliminating sinusoidal disturbances in communication and others. The second-order adaptive notch filters have been used for the elimination or enhancement of sine wave components with unknown frequencies[ I]-[7]. For adaptive IIR notch filter (Am) analysis, a good algorithm requires fast convergence speed and low variance and bias [I]. One approaches of using the adaptive algorithm of quantize least mean p-power error criterion (QLMp) and variable step-size quantize least mean p-power error criterion (VSQLMp) were proposed in references [3][4], respectively. However, this algorithm performs high variance.In this paper, we have modified the algorithm by using the variable step-size was proposed in reference [4]. The new algorithm increases the noise robustness and reduces the computational complexity when compare with the previous one. Assuming a sinusoidal signal with Gaussian white noise as input signal of the second-order adaptive IIR notch filter, the filter is analyzed for various 0-7803-7690-0/02/$17.00 02002 IEEE where A is the amplitude, $ is the 'phase, WO is an unknown frequency, n is time index and v(n) is the zero mean unit variance white Gaussian noise, the input signal to noise ratio (SM) is in the S m (dB) = lolog A2/2O2 form ( O2 = noise variance). The objective is to estimate unknown frequency, U. by applying n(n) to the ANF, the transfer function of the second-order adaptive IIR notch filter is given by: 1 + a(n)z-' + z-2 1 + pa(r2)z-l + p2z-2(2)
H ( z ) =where a@) is the...