I am streaming accelerometer data from my android phone and have successfully built a live plot using matplotlib. I am using the comma operator to dynamically update the plot but I am wondering if there is a more elegant/pythonic way to do it. To execute the code below you must use the app Sensorstream IMU+GPS. The code below will grab the accelerometer values and plot them live. I based the plotting on Can you plot live data in matplotlib?. Like I said it works but the code is clumsy. Even with speedups mentioned in the matplotlib documentation I am running at about 25 FPS. The technique, if I only use a simple plot can get up to about 90 FPS. It can be shown that you can achieve ~200 FPS of faster here why is plotting with Matplotlib so slow?. I cannot find my bottleneck. So, is there a more elegant way to code up all the subplots? Second, can I speed up the plotting?
import socket, traceback
import matplotlib
matplotlib.use('TkAgg')
import matplotlib.pyplot as plt
from scipy.signal import butter, lfilter,iirfilter,savgol_filter
import math
import pylab
from pylab import *
import time
import numpy as np
host = ''
port = 5555
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
s.setsockopt(socket.SOL_SOCKET, socket.SO_BROADCAST, 1)
s.bind((host, port))
# lists for plotting
Ax = [0.0] * 50
Ay = [0.0] * 50
Az = [0.0] * 50
G = [0.0] * 50
x = [i for i in range(len(Ax))]
#used for debugging
fig = plt.figure(figsize=(16,10))
# raw data
ax = plt.subplot("311")
ax.set_xlim(0, 50)
ax.set_ylim(-2, 2)
ax.set_title("Raw acceleration data")
ax.set_ylabel("g$/m^2$",fontsize=18)
ax.hold(True)
line = ax.plot(Ax,label='Acc x')[0]
line2 = ax.plot(Ay,label='Acc y')[0]
line3 = ax.plot(Az,label='Acc z')[0]
# filtered data
ax2 = plt.subplot("312")
ax2.set_xlim(0, 50)
ax2.set_ylim(-2, 2)
ax2.set_title(" acceleration data")
ax2.set_ylabel("g$/m^2$",fontsize=18)
ax2.hold(True)
f_line = ax2.plot(Ax,label='Acc x')[0]
f_line2 = ax2.plot(Ay,label='Acc y')[0]
f_line3 = ax2.plot(Az,label='Acc z')[0]
# tilt angle plot
ax3 = plt.subplot("313")
ax3.set_ylim([-180,180])
ax3.set_title("Tilt Angles")
ax3.set_ylabel("degrees",fontsize=18)
t_line = ax3.plot(G)[0]
fig.suptitle('Three-axis accelerometer streamed from Sensorstream',fontsize=18)
plt.show(False)
plt.draw()
# cache the background
background = fig.canvas.copy_from_bbox(fig.bbox)
count = 0
print("Success binding")
while 1:
# time it
tstart = time.time()
message, address = s.recvfrom(8192)
messageString = message.decode("utf-8")
Acc = messageString.split(',')[2:5]
Acc = [float(Acc[i])/10.0 for i in range(3)]
# appending and deleting is order 10e-5 sec
Ax.append(Acc[0])
del Ax[0]
Ay.append(Acc[1])
del Ay[0]
Az.append(Acc[2])
del Az[0]
G.append(np.sqrt(Ax[-1]**2 + Ay[-1]**2 + Az[-1]**2))
del G[0]
# filter
acc_x_savgol = savgol_filter(Ax, window_length=5, polyorder=3)
acc_y_savgol = savgol_filter(Ay, window_length=5, polyorder=3)
acc_z_savgol = savgol_filter(Az, window_length=5, polyorder=3)
tilt_angles = []
for i,val in enumerate(G):
angle = math.atan2(Ax[i], -1*Ay[i]) * (180 / math.pi)
if (math.isnan(angle)):
tilt_angles.append(0)
else:
tilt_angles.append(angle)
print(Ax[0],Ay[1],Az[2])
line.set_xdata(x)
line.set_ydata(Ax)
line2.set_xdata(x)
line2.set_ydata(Ay)
line3.set_xdata(x)
line3.set_ydata(Az)
ax.set_xlim(count, count+50)
f_line.set_xdata(x)
f_line.set_ydata(acc_x_savgol)
f_line2.set_xdata(x)
f_line2.set_ydata(acc_y_savgol)
f_line3.set_xdata(x)
f_line3.set_ydata(acc_z_savgol)
ax2.set_xlim(count, count+50)
t_line.set_xdata(x)
t_line.set_ydata(tilt_angles)
ax3.set_xlim(count, count+50)
# restore background
fig.canvas.restore_region(background)
# redraw just the points
ax.draw_artist(line)
ax.draw_artist(line2)
ax.draw_artist(line3)
ax2.draw_artist(f_line)
ax2.draw_artist(f_line2)
ax2.draw_artist(f_line3)
ax3.draw_artist(t_line)
# fill in the axes rectangle
fig.canvas.blit(fig.bbox)
count+=1
x = np.arange(count,count+50,1)
# tops out at about 25 fps :|
print "Total time for 1 plot is: ",(time.time() - tstart)