中文摘要：

小波分解尺度选择是小波去噪中的关键，分解尺度过大或过小，都会直接影响传感信号去噪的质量。本文通过分析常规的分解尺度选择方法，引用指标融合算法原理从图形和统计两个角度出发选用信噪比（SNR）、均方根误差（RMSE）、平滑度r以及信号偏差（BIAS）4项评价指标融合成一项指标，采用指数函数对融合指标进行拟合，并提出一种拐点判别公式来准确判定最优分解尺度。然后，采用所提方法，结合改进的阈值和阈值函数，对实验信号进行最佳分解尺度寻优，实验验证结果表明，本文方法所确定的最优分解尺度对应的中心波长反射率与实际值相符，因此能够获得更优的去噪效果。

英文摘要：

In demodulating process of optical fiber sensing, the existence of the noise is inevitable, seriously affecting system＇s precision to demodulate central wavelength. The wavelet de-noising method with its good ability of time-frequency localization, multi-resolution analysis and handling nonlinear problem in the field of optical fiber sensing de-noising has been widely used,and it is crucial to select the optimization decomposition scale, since the decomposition scale will affect the quality of denoising directly. Based on the analysis of conventional and proposed methods of decomposition scale selection, a multi-indicator fusing method is quoted to fuse signal to noise ratio, root mean square error, smoothness and bias of signal into a comprehensive evaluating indicator, and process the fusion indicator with exponential fitting algorithm and present an inflection point judgment formula to determine the optimal decomposition scale. And then, combined with better threshold and improved threshold function, with experimental signal, the best decomposition scale is chosen in this presented method. The experimental results show that the reflectivity rate of central wavelength, with optimal decomposition scale selected in this method, is consistent with the real values. Thus, the proposed method shows better performance in de-noising.