基于损失函数的采样策略研究

Research on sampling strategy based on loss function

使用ADC芯片进行数据转换时,可在过采样条件下用算术平均滤波法降噪,但该算法在噪声RMS值可变系统中难以兼顾效率和降噪效果。基于损失函数对算术平均滤波法做出了改进,将平均次数与噪声RMS和分辨率(信噪比)相对应,再根据分辨率(信噪比)要求,对不同的噪声RMS值(信噪比)使用不同的过采样次数。理论分析和实测结果表明,该方法可在更大噪声RMS值变化范围内获得更好的降噪结果和更高的采样效率。

When the ADC chip is used for data conversion,it is possible to use the arithmetic average filtering method to reduce noise by oversampling.However,it is difficult to balance the efficiency and the noise reduction effect in a system with variable noise RMS.Based on the loss function,the arithmetic average filtering method is improved,and the average number is corresponding to the noise RMS and resolution(signal-to-noise ratio).According to the requirements of resolution(signal-to-noise ratio),different noise RMS values(signal-to-noise ratio)Apply different averages.Theoretical analysis and measured results show that this method can obtain better noise reduction results and higher sampling efficiency in a larger range of noise RMS.