1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
|
"""
This module provides functions for matching a pattern within a trace to it.
"""
import numpy as np
from scipy.signal import find_peaks
from public import public
from typing import List
from .process import normalize
from .edit import trim
from .trace import Trace
@public
def match_pattern(trace: Trace, pattern: Trace, threshold: float = 0.8) -> List[int]:
normalized = normalize(trace)
pattern_samples = normalize(pattern).samples
correlation = np.correlate(normalized.samples, pattern_samples, "same")
correlation = (correlation - np.mean(correlation)) / (np.max(correlation))
peaks, props = find_peaks(correlation, prominence=(threshold, None))
pairs = sorted(zip(peaks, props["prominences"]), key=lambda it: it[1], reverse=True)
half = len(pattern_samples) // 2
filtered_peaks: List[int] = []
for peak, prominence in pairs:
if not filtered_peaks:
filtered_peaks.append(peak - half)
else:
for other_peak in filtered_peaks:
if abs((peak - half) - other_peak) <= len(pattern_samples):
break
else:
filtered_peaks.append(peak - half)
return filtered_peaks
@public
def match_part(trace: Trace, offset: int, length: int) -> List[int]:
return match_pattern(trace, trim(trace, offset, offset + length))
|
