如何提取山峰？

Question

在此之前，我分析了信号的FFT，并成功地提取了峰值。

信号看上去如下：

​

​

上面的信号是用峰值提取的：

import peakutils

peaks = peakutils.indexes(signal_fft, thres=0.2, min_dist=5)

我试图以同样的方式从下面的变换中提取出峰值：

​

​

然而，对于这样的信号，我还没有找到一个阈值，只有峰值才能在没有噪音的情况下被提取。

原始转换如下：

signal_fft_new = [
...
45.59,0.0036787
45.674,0.0037884
45.758,0.0039009
45.842,0.0039426
45.926,0.0039838
46.01,0.0041179
46.095,0.0041863
46.179,0.0042501
46.263,0.004359
46.347,0.0044381
46.431,0.0045555
46.515,0.0046992
46.599,0.0047373
46.683,0.0048359
46.767,0.0050033
46.852,0.0051318
46.936,0.0052597
47.02,0.0054131
47.104,0.0055889
47.188,0.0057262
47.272,0.0058769
47.356,0.0060621
47.44,0.0062528
47.524,0.00647
47.609,0.0067256
47.693,0.0068882
47.777,0.0071613
47.861,0.0074998
47.945,0.0077778
48.029,0.008125
48.113,0.0084604
48.197,0.0087574
48.281,0.0092132
48.366,0.0096537
48.45,0.010139
48.534,0.010737
48.618,0.011321
48.702,0.012016
48.786,0.012889
48.87,0.013849
48.954,0.014952
49.039,0.015997
49.123,0.017455
49.207,0.019221
49.291,0.021306
49.375,0.02449
49.459,0.027053
49.543,0.031536
49.627,0.039624
49.711,0.045848
49.796,0.067003
49.88,0.11004
49.964,0.14256
50.048,0.32163
50.132,0.38476
50.216,0.14082
50.3,0.074299
50.384,0.047628
50.468,0.040767
50.553,0.032828
50.637,0.027988
50.721,0.024559
50.805,0.021452
50.889,0.019469
50.973,0.017578
51.057,0.016165
51.141,0.015044
51.225,0.01374
51.31,0.012839
51.394,0.012016
51.478,0.011213
51.562,0.010747
51.646,0.01011
51.73,0.0095173
51.814,0.0091049
51.898,0.0087235
...
]

我试图通过以下方式提取这些山峰：

signal = np.genfromtxt(file_path, delimiter=',')
signal[:, 1] = np.nan_to_num(np.log(signal[:, 1]))
peaks = peakutils.indexes(signal[:, 1], thres=0.2, min_dist=5)

有谁知道如何排除噪音才能得到峰值指数呢？

Answer 1

老式的方法是：

trend.

• Determine

1. 低通滤波器(平均5个相邻的桶)得到一个趋势线；减去噪声
2. 的标准差，选择2个标准差以上的值。