小波在时域、频域都具有表征信号局部特征的能力,对于语音数据的去噪,与其他方法比较具有优势,但是阈值的选择缺乏明确的指标。模极大值法保留幅度随尺度增加而增大的点(对应于有用信号的极值点)来达到去噪的目的,但将非极值处变换系数置零,造成信号损失。提出了一种变化模极大值与阈值决策相融合的去噪方法,有效地结合了两者的优点。经实验证明,该方法精度高、算法简单、计算速度快。
This paper proposed a new wavelet-denoising method, which was based on wavelet transform modulus maximum and nonlinear threshold denoising method on wavelet. This method absorbed the advantages and overcomed the disadvantages of traditional methods. Take an experiment to validate the improved method, which shows the new method has the advantage of the high precision and the simple arithmetic and the high computing efficiency.