Gamma spectrometry is a technique used for measuring the quantitative and qualitative determination of radioisotopes emitting gamma radiation. Each element of the electronic measurement chain has its own limitations, and therefore may introduce noise in the recorded signal and thus make it difficult to interpret. In particular, the noisy and random nature of the output signal of the preamplifier, and the pileup phenomenon remain major problems in the field of gamma spectrometry signal processing. For a better use of spectrometric signals, we propose in this work a method for adaptive filtering of the preamplifier output signal based on least mean-square algorithm (LMS). This choice is justified because of the simplicity and relevance of this method which is widely used in similar applications. The use of high performance data acquisition card and the implementation of this method on digital signal processor (DSP) allow real time processing of the recorded signals. Various tests were performed using different signals as reference input of the filter, both in the presence and absence of a gamma radiation source. Preliminary results show a significant improvement of the signal to noise ratio (SNR), especially when the reference input is chosen to be the recorded background noise of the detector.