Abstract: Nonlinear regression function obtained by making fully use of wavelet transform is proposed.The regression parameters of signal are gained by reconstructing wavelet coefficients after wavelet coefficients of detail part are influenced by threshold during wavelet decomposition of using wavelet to transform the signal into approximate part and detail part.An algorithm is described and an example is illustrated to this method here.The PSNR of signal gained by the method is lower，and the signal is smoother，so the method is effective and feasible.

Key words: wavelet transform, nonlinear regression, threshold, de-noising

摘要： 提出了一种利用小波变换得到非线性回归估计函数的方法，把信号分解成近似部分和细节部分，对细节部分的小波系数进行阈值去噪，然后通过重构该函数得到信号。给出了基于小波变换的非线性回归分析算法，并进行了仿真实验，该方法得到的回归信号信噪比更低，更光滑，表明该方法是有效的。

关键词: 小波变换, 非线性回归, 阈值, 去噪