信号压缩

时间:2013-03-27 12:47:15

标签: python signal-processing python-2.5

我需要“压缩”代表信号的python数组的大小。信号如下图所示。

signal = [
    [0.0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0,1.1], #time values
    [1,1,1,2,3,4,4,4,4,2,1,1] #function values
    ]

压缩后,信号应如下面的代码所示。

signal_compressed = [
    [0.0,0.2,0.3,0.4,0.5,0.8,0.9,1.0,1.1], #time values
    [1,1,2,3,4,4,2,1,1] #function values
    ]

您会看到,如果存在具有常量值的区域,则仅存储此区域的第一个和最后一个值。

我写了以下算法来做到这一点。

signal_compressed = [[],[]]

old_value = None
for index, value in enumerate(signal[1]):
    if value != old_value:
        if index > 0:
            if signal_compressed[0][-1] != signal[0][index - 1]:
                signal_compressed[0].append(signal[0][index - 1])
                signal_compressed[1].append(signal[1][index - 1])
        signal_compressed[0].append(signal[0][index])
        signal_compressed[1].append(value)
        old_value = value

if signal_compressed[0][-1] < signal[0][-1]:
    signal_compressed[0].append(signal[0][-1])
    signal_compressed[1].append(signal[1][-1])

此算法运行正常。对于具有大量恒定段的信号,他的工作速度非常快。但是,如果我尝试压缩没有恒定段的信号(例如正弦信号或噪声信号),算法的工作速度非常慢。

如何加速算法并保存功能?

2 个答案:

答案 0 :(得分:1)

以下是使用生成器执行此操作的一种方法:

def compress(signal):
    prev_t, prev_val = None, None
    for t, val in zip(*signal):
        if val != prev_val:
            if prev_t is not None:
                yield prev_t, prev_val
            yield t, val
            prev_t, prev_val = None, val
        else:
            prev_t, prev_val = t, val
    if prev_t is not None:
        yield prev_t, prev_val

signal = [
    [0.0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0,1.1], #time values
    [1,1,1,2,3,4,4,4,4,2,1,1] #function values
    ]
print zip(*compress(signal))

我认为转换signal会更自然,就像这样存储它:

[(0.0, 1),
 (0.1, 1),
 (0.2, 1),
 (0.3, 2),
 (0.4, 3),
 (0.5, 4),
 (0.6, 4),
 (0.7, 4),
 (0.8, 4),
 (0.9, 2),
 (1.0, 1),
 (1.1, 1)]

这样两个zip(*seq)调用就没必要了,整个处理都可以动态完成。

最后,如果对于大输入来说这仍然太慢,那么可能值得研究使用NumPy。以下是一个这样的解决方案的概述:

import numpy as np

signal = [
    [0.0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0,1.1], #time values
    [1,1,1,2,3,4,4,4,4,2,1,1] #function values
    ]

def npcompress(signal):
    sig=np.array(signal)
    idx = np.where(sig[1][1:] != sig[1][:-1])[0]
    idx_arr = np.sort(np.array(list(set(idx) | set(idx + 1) | set([0]) | set([len(sig[1]) - 1]))))
    return sig.T[idx_arr]

print npcompress(signal).T

答案 1 :(得分:1)

您可以使用itertools.groupby()

In [93]: from itertools import groupby

In [94]: from operator import itemgetter

In [95]: ti=[0.0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0,1.1]

In [96]: fun=[1,1,1,2,3,4,4,4,4,2,1,1]

In [97]: def func(ti,func):
    new_time=[]
    new_func=[]
    for k,v in groupby(enumerate(func),itemgetter(1)):
       lis=list(v)
       if len(lis)>1:
           ind1,ind2=lis[0][0],lis[-1][0]
           new_time.extend([ti[ind1],ti[ind2]])
           new_func.extend([func[ind1],func[ind2]])
       else:    
           new_time.append(ti[lis[0][0]])
           new_func.append(func[lis[0][0]])
    return new_time,new_func
   ....: 

In [98]: func(ti,fun)
Out[98]: ([0.0, 0.2, 0.3, 0.4, 0.5, 0.8, 0.9, 1.0, 1.1],
          [1, 1, 2, 3, 4, 4, 2, 1, 1])