Abstract: In this paper, the finite word-length effects in the normalized least mean square (LMS) adaptive filter is discussed, and the factors influencing the quantized mean square error (MSE) of the normalized LMS adaptive filter are analyzed. To ensure that the finite word-length effects of the normalized LMS adaptive filter meet performance requirements in target detection, a method combining statistical analysis of MSE with target detection performance analysis is proposed for determining the operational word length of the filter. The relationship of the step size, filter order with the MSE of the normalized LMS adaptive filter and with the target detection signal-to-noise ratio (SNR), is examined. Meanwhile, the correlations of the variances introduced by parameters such as the filter output word length, filter input word length, and filter weight update word length, with the MSE and the target detection SNR of the normalized LMS adaptive filter, are analyzed. Simulation experiments are conducted to validate the analytical results. The simulation results demonstrate that the finite word-length adaptive filter designed using this method achieves the minimum quantized MSE and the maximum target detection SNR, exhibiting optimal filtering performance in narrow sense.