Added DTW to the PythonWrapper API
parent
b46a2f86b5
commit
82559adce2
@ -0,0 +1,154 @@
|
||||
# Bug corrections for version 1.9
|
||||
import cmsisdsp as dsp
|
||||
import cmsisdsp.fixedpoint as f
|
||||
import numpy as np
|
||||
import colorama
|
||||
from colorama import init,Fore, Back, Style
|
||||
from numpy.testing import assert_allclose
|
||||
|
||||
from numpy.linalg import norm
|
||||
from dtw import *
|
||||
|
||||
import matplotlib
|
||||
import matplotlib as mpl
|
||||
import matplotlib.pyplot as plt
|
||||
|
||||
init()
|
||||
|
||||
def printTitle(s):
|
||||
print("\n" + Fore.GREEN + Style.BRIGHT + s + Style.RESET_ALL)
|
||||
|
||||
def printSubTitle(s):
|
||||
print("\n" + Style.BRIGHT + s + Style.RESET_ALL)
|
||||
|
||||
printTitle("DTW Window")
|
||||
|
||||
printSubTitle("SAKOE_CHIBA_WINDOW")
|
||||
|
||||
refWin1=np.array([[1, 1, 1, 0, 0],
|
||||
[1, 1, 1, 1, 0],
|
||||
[1, 1, 1, 1, 1],
|
||||
[0, 1, 1, 1, 1],
|
||||
[0, 0, 1, 1, 1],
|
||||
[0, 0, 0, 1, 1],
|
||||
[0, 0, 0, 0, 1],
|
||||
[0, 0, 0, 0, 0],
|
||||
[0, 0, 0, 0, 0],
|
||||
[0, 0, 0, 0, 0]], dtype=np.int8)
|
||||
|
||||
dtwWindow=np.zeros((10,5),dtype=np.int8)
|
||||
wsize=2
|
||||
status,w=dsp.arm_dtw_init_window_q7(dsp.ARM_DTW_SAKOE_CHIBA_WINDOW,wsize,dtwWindow)
|
||||
|
||||
assert (w==refWin1).all()
|
||||
|
||||
printSubTitle("SLANTED_BAND_WINDOW")
|
||||
|
||||
refWin2=np.array([[1, 1, 0, 0, 0],
|
||||
[1, 1, 0, 0, 0],
|
||||
[1, 1, 1, 0, 0],
|
||||
[0, 1, 1, 0, 0],
|
||||
[0, 1, 1, 1, 0],
|
||||
[0, 0, 1, 1, 0],
|
||||
[0, 0, 1, 1, 1],
|
||||
[0, 0, 0, 1, 1],
|
||||
[0, 0, 0, 1, 1],
|
||||
[0, 0, 0, 0, 1]], dtype=np.int8)
|
||||
|
||||
dtwWindow=np.zeros((10,5),dtype=np.int8)
|
||||
wsize=1
|
||||
status,w=dsp.arm_dtw_init_window_q7(dsp.ARM_DTW_SLANTED_BAND_WINDOW,wsize,dtwWindow)
|
||||
|
||||
assert (w==refWin2).all()
|
||||
|
||||
|
||||
printTitle("DTW Cost Matrix and DTW Distance")
|
||||
|
||||
QUERY_LENGTH = 10
|
||||
TEMPLATE_LENGTH = 5
|
||||
|
||||
query=np.array([ 0.08387197, 0.68082274, 1.06756417, 0.88914541, 0.42513398, -0.3259053,
|
||||
-0.80934885, -0.90979435, -0.64026483, 0.06923695])
|
||||
|
||||
template=np.array([ 1.00000000e+00, 7.96326711e-04, -9.99998732e-01, -2.38897811e-03,
|
||||
9.99994927e-01])
|
||||
|
||||
cols=np.array([1,2,3])
|
||||
rows=np.array([10,11,12])
|
||||
|
||||
printSubTitle("Without a window")
|
||||
|
||||
referenceCost=np.array([[0.91612804, 0.9992037 , 2.0830743 , 2.1693354 , 3.0854583 ],
|
||||
[1.2353053 , 1.6792301 , 3.3600516 , 2.8525472 , 2.8076797 ],
|
||||
[1.3028694 , 2.3696373 , 4.4372 , 3.9225004 , 2.875249 ],
|
||||
[1.4137241 , 2.302073 , 4.1912174 , 4.814035 , 2.9860985 ],
|
||||
[1.98859 , 2.2623994 , 3.6875322 , 4.115055 , 3.5609593 ],
|
||||
[3.3144953 , 2.589101 , 3.2631946 , 3.586711 , 4.8868594 ],
|
||||
[5.123844 , 3.3992462 , 2.9704008 , 3.7773607 , 5.5867043 ],
|
||||
[7.0336385 , 4.309837 , 3.0606053 , 3.9680107 , 5.8778 ],
|
||||
[8.673903 , 4.950898 , 3.420339 , 4.058215 , 5.698475 ],
|
||||
[9.604667 , 5.0193386 , 4.489575 , 3.563591 , 4.494349 ]],
|
||||
dtype=np.float32)
|
||||
|
||||
referenceDistance = 0.2996232807636261
|
||||
|
||||
# Each row is a new query
|
||||
a,b = np.meshgrid(template,query)
|
||||
distance=abs(a-b).astype(np.float32)
|
||||
|
||||
status,dtwDistance,dtwMatrix = dsp.arm_dtw_distance_f32(distance,None)
|
||||
|
||||
|
||||
assert_allclose(referenceDistance,dtwDistance)
|
||||
assert_allclose(referenceCost,dtwMatrix)
|
||||
|
||||
printSubTitle("Path")
|
||||
|
||||
path=dsp.arm_dtw_path_f32(np.copy(dtwMatrix))
|
||||
#print(path)
|
||||
pathMatrix=np.zeros(dtwMatrix.shape)
|
||||
for x in list(zip(path[0::2],path[1::2])):
|
||||
pathMatrix[x] = 1
|
||||
|
||||
|
||||
fig, ax = plt.subplots()
|
||||
im = ax.imshow(pathMatrix,vmax=2.0)
|
||||
|
||||
for i in range(QUERY_LENGTH):
|
||||
for j in range(TEMPLATE_LENGTH):
|
||||
text = ax.text(j, i, "%.1f" % dtwMatrix[i, j],
|
||||
ha="center", va="center", color="w")
|
||||
fig.tight_layout()
|
||||
plt.show()
|
||||
|
||||
printSubTitle("With a window")
|
||||
|
||||
referenceDistance = 0.617099940776825
|
||||
referenceCost=np.array([[9.1612804e-01, 9.9920368e-01, np.NAN, np.NAN,
|
||||
np.NAN],
|
||||
[1.2353053e+00, 1.6792301e+00, np.NAN, np.NAN,
|
||||
np.NAN],
|
||||
[1.3028694e+00, 2.3696373e+00, 4.4372001e+00, np.NAN,
|
||||
np.NAN],
|
||||
[np.NAN, 3.0795674e+00, 4.9687119e+00, np.NAN,
|
||||
np.NAN],
|
||||
[np.NAN, 3.5039051e+00, 4.9290380e+00, 5.3565612e+00,
|
||||
np.NAN],
|
||||
[np.NAN, np.NAN, 4.8520918e+00, 5.1756082e+00,
|
||||
np.NAN],
|
||||
[np.NAN, np.NAN, 5.0427418e+00, 5.8497019e+00,
|
||||
7.6590457e+00],
|
||||
[np.NAN, np.NAN, np.NAN, 6.7571073e+00,
|
||||
8.6668968e+00],
|
||||
[np.NAN, np.NAN, np.NAN, 7.3949833e+00,
|
||||
9.0352430e+00],
|
||||
[np.NAN, np.NAN, np.NAN, np.NAN,
|
||||
9.2564993e+00]], dtype=np.float32)
|
||||
|
||||
|
||||
status,dtwDistance,dtwMatrix = dsp.arm_dtw_distance_f32(distance,w)
|
||||
|
||||
|
||||
assert_allclose(referenceDistance,dtwDistance)
|
||||
assert_allclose(referenceCost[w==1],dtwMatrix[w==1])
|
||||
|
||||
Loading…
Reference in New Issue