A conventional single-channel anti-Stokes temperature measurement system usually demands a considerately long measuring period. To overcome such a drawback while guarantee time and space resolution of the system, a method was proposed in this paper, which calls for a much shorter period by employing modern DSP such as wavelet decomposition, multi-resolution analysis and wavelet reconstruction .The characteristics of distributed temperature measurement signal were discussed first. Then, how to select proper wavelet functions was emphasized as well as how to ascertain the levels of wavelet decomposition by means of Shannon entropy.The experiments indicate that the measurement period is greatly reduced, thus the system becomes more efficient.