Multiple Spectrograms in the differents plots.
118 views
Skip to first unread message
Aldrich Cabrera
Dec 16, 2019, 1:53:42 PM12/16/19
to pyqtgraph
Hi guys,
I'm trying to plot in real-time several spectrograms in different plots. When I plot only one channel works great but after adding the others channels the result is shown in the same plot and I'd want to show the result by channel in different plots.
I'm getting the data from a thread. and the code I'm using to plot looks like this:
import numpy as np
import pyqtgraph as pg
import pyqtgraph.exporters
import pyaudio
from PyQt4 import QtCore, QtGui
import cv2
FS = 48000 #Hz
CHUNKSZ = 1024 #samples
class MicrophoneRecorder():
def __init__(self, signal):
self.signal = signal
#self.signal2 = signal2
self.p = pyaudio.PyAudio() # Create pyaudio instantiation
self.stream = self.p.open(format=pyaudio.paInt16,
rate=FS,
channels=8,
input_device_index=4,
input=True,
frames_per_buffer=CHUNKSZ)
def read(self):
self.stream.start_stream()
data = self.stream.read(CHUNKSZ)
y = np.fromstring(data, dtype=np.int16)
yC1 = y[0::8] # Channel 1
yC2 = y[1::8] # Channel 2
yC3 = y[2::8] # Channel 3
yC4 = y[3::8] # Channel 4
yC5 = y[4::8] # Channel 5
yC6 = y[5::8] # Channel 6
yC7 = y[6::8] # Channel 7
yC8 = y[7::8] # Channel 8
self.signal.emit(yC1)
self.signal.emit(yC2)
def close(self):
self.stream.stop_stream()
self.stream.close()
self.p.terminate()
class SpectrogramWidget(pg.PlotWidget):
read_collected = QtCore.pyqtSignal(np.ndarray)
#print(read_collected)
#read_collected2 = QtCore.pyqtSignal(np.ndarray)
def __init__(self):
super(SpectrogramWidget, self).__init__(parent=None, background='default')
self.img = pg.ImageItem() # Displaying an image
self.img2 = pg.ImageItem() # Displaying an image
self.addItem(self.img) # Add a graphics item to the view box
self.addItem(self.img2) # Add a graphics item to the view box
#self.img_array = np.zeros((1000, CHUNKSZ/2+1))
self.img_array = np.zeros((94, CHUNKSZ/2+1)) # Return new array
print(self.img_array)
# bipolar colormap
pos = np.array([0., 1., 0.5, 0.25, 0.75]) # Object, dtype, copy, order, subok, ndmin
color = np.array([[0,255,255,255], [255,255,0,255], [0,0,0,255], (0, 0, 255, 255), (255, 0, 0, 255)], dtype=np.ubyte)
cmap = pg.ColorMap(pos, color) # pos, color - relationship between scalar value and range of color
lut = cmap.getLookupTable(0.0, 1.0, 256) # Start value, final value, number of points
self.img.setLookupTable(lut) # Accepts currents images
self.img.setLevels([-50,40]) # Blacklevel, Whitelevel
freq = np.arange((CHUNKSZ/2)+1)/(float(CHUNKSZ)/FS) # Spaced values within a given interval
yscale = 1.0/(self.img_array.shape[1]/freq[-1]) # Scale the y-axis for intervals of freq
self.img.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.win = np.hanning(CHUNKSZ) # Return Hanning Window
self.show() # Show the image generated
def update(self, chunk):
# normalized, windowed frequencies in data chunk
spec = np.fft.rfft(chunk*self.win) / CHUNKSZ # Compute discrete FT for real input
# get magnitude
psd = abs(spec)
# convert to dB scale
psd = 20 * np.log10(psd)
# roll down one and replace leading edge with new data
self.img_array = np.roll(self.img_array, -1, 0)
self.img_array[-1:] = psd
self.img.setImage(self.img_array, autoLevels=False)
#self.imagewindow.clear()
#self.imagewindow.addItem(self.img)
#exporter = pg.exporters.ImageExporter(self.img_array)
#exporter.parameters()['width'] = 100
#exporter.export('test.png')
if __name__ == '__main__':
app = QtGui.QApplication([])
w = SpectrogramWidget()
w.read_collected.connect(w.update)
#w.read_collected2.connect(w.update)
mic = MicrophoneRecorder(w.read_collected)
# time (seconds) between reads
interval = FS/CHUNKSZ
t = QtCore.QTimer()
t.timeout.connect(mic.read)
t.start(100/interval) #QTimer takes ms
app.exec_()
mic.close()
I appreciate any suggestions.
Grant R
Dec 16, 2019, 10:19:14 PM12/16/19
to pyqtgraph
Are you trying to plot eight different images in eight different plot views? In your update method you are only setting the image for self.img which is the one image view. One way is to have eight different image views and pass the index of the one you want to update via your signal. If this is your layout, then I think you should use a GraphicsLayoutWidget with a GraphicsLayout and your items inside (http://www.pyqtgraph.org/documentation/plotting.html).
Aldrich Cabrera
Dec 17, 2019, 1:30:06 PM12/17/19
to pyqtgraph
Hi Grant R,
Yeah, I'm trying to plot eight different images. Thanks for the information was useful.
Already I combination these eight views using "GraphicsLayoutWidget" in a single plot, but about the index in the signal to show the right channel, I'm confused. I have created an image (Attached image) with eight plot, but it shows the same channel in the eight views, and I want to show each channel by separate.
I apologise for the inconvenience, but I'm new in this and is very complicated to me. The design was taken: https://gist.github.com/boylea/1a0b5442171f9afbf372.
And my update method thus looks :
class SpectrogramWidget(pg.PlotWidget):
read_collected = QtCore.pyqtSignal(np.ndarray)
read_collected2 = QtCore.pyqtSignal(np.ndarray)
def __init__(self):
super(SpectrogramWidget, self).__init__(parent=None, background='default')
self.win = pg.GraphicsLayoutWidget(show=True, title="Spectrograms")
self.win.resize(1280,720)
self.win.setWindowTitle('Spectrograms')
self.p1 = self.win.addPlot(title="Channel 1")
self.p2 = self.win.addPlot(title="Channel 2")
self.p3 = self.win.addPlot(title="Channel 3")
self.win.nextRow()
self.p4 = self.win.addPlot(title="Channel 4")
self.p5 = self.win.addPlot(title="Channel 5")
self.p6 = self.win.addPlot(title="Channel 6")
self.win.nextRow()
self.p7 = self.win.addPlot(title="Channel 7")
self.p8 = self.win.addPlot(title="Channel 8")
self.img = pg.ImageItem() # Displaying an image
self.img2 = pg.ImageItem() # Displaying an image
self.img3 = pg.ImageItem() # Displaying an image
self.img4 = pg.ImageItem() # Displaying an image
self.img5 = pg.ImageItem() # Displaying an image
self.img6 = pg.ImageItem() # Displaying an image
self.img7 = pg.ImageItem() # Displaying an image
self.img8 = pg.ImageItem() # Displaying an image
self.p1.addItem(self.img) # Add a graphics item to the view box
self.p2.addItem(self.img2) # Add a graphics item to the view box
self.p3.addItem(self.img3) # Add a graphics item to the view box
self.p4.addItem(self.img4) # Add a graphics item to the view box
self.p5.addItem(self.img5) # Add a graphics item to the view box
self.p6.addItem(self.img6) # Add a graphics item to the view box
self.p7.addItem(self.img7) # Add a graphics item to the view box
self.p8.addItem(self.img8) # Add a graphics item to the view box
#self.img_array = np.zeros((1000, CHUNKSZ/2+1))
self.img_array = np.zeros((94, CHUNKSZ/2+1)) # Return new array
# bipolar colormap
pos = np.array([0., 1., 0.5, 0.25, 0.75]) # Object, dtype, copy, order, subok, ndmin
color = np.array([[0,255,255,255], [255,255,0,255], [0,0,0,255], (0, 0, 255, 255), (255, 0, 0, 255)], dtype=np.ubyte)
cmap = pg.ColorMap(pos, color) # pos, color - relationship between scalar value and range of color
lut = cmap.getLookupTable(0.0, 1.0, 256) # Start value, final value, number of points
self.img.setLookupTable(lut) # Accepts currents images
self.img.setLevels([-50,40]) # Blacklevel, Whitelevel
self.img2.setLookupTable(lut) # Accepts currents images
self.img2.setLevels([-50,40]) # Blacklevel, Whitelevel
self.img3.setLookupTable(lut) # Accepts currents images
self.img3.setLevels([-50,40]) # Blacklevel, Whitelevel
self.img4.setLookupTable(lut) # Accepts currents images
self.img4.setLevels([-50,40]) # Blacklevel, Whitelevel
self.img5.setLookupTable(lut) # Accepts currents images
self.img5.setLevels([-50,40]) # Blacklevel, Whitelevel
self.img6.setLookupTable(lut) # Accepts currents images
self.img6.setLevels([-50,40]) # Blacklevel, Whitelevel
self.img7.setLookupTable(lut) # Accepts currents images
self.img7.setLevels([-50,40]) # Blacklevel, Whitelevel
self.img8.setLookupTable(lut) # Accepts currents images
self.img8.setLevels([-50,40]) # Blacklevel, Whitelevel
freq = np.arange((CHUNKSZ/2)+1)/(float(CHUNKSZ)/FS) # Spaced values within a given interval
yscale = 1.0/(self.img_array.shape[1]/freq[-1]) # Scale the y-axis for intervals of freq
self.img.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.img2.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.img3.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.img4.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.img5.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.img6.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.img7.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.img8.scale((1./FS)*CHUNKSZ, yscale) # Scale between the time and freq
self.p1.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.p1.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.p2.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.p2.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.p3.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.p3.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.p4.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.p4.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.p5.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.p5.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.p6.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.p6.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.p7.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.p7.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.p8.setLabel('left', 'Frequency', units='Hz') # Set y-axi label
self.p8.setLabel('bottom', 'Time', units='sec') # Set x-axi label
self.win2 = np.hanning(CHUNKSZ) # Return Hanning Window
#print(self.win)
self.p1.show() # Show the image generated
def update(self, chunk):
# normalized, windowed frequencies in data chunk
spec = np.fft.rfft(chunk*self.win2) / CHUNKSZ # Compute discrete FT for real input
#print(spec)
# get magnitude
psd = abs(spec)
# convert to dB scale
psd = 20 * np.log10(psd)
# roll down one and replace leading edge with new data
self.img_array = np.roll(self.img_array, -1, 0)
self.img_array[-1:] = psd
self.img.setImage(self.img_array, autoLevels=False)
self.img2.setImage(self.img_array, autoLevels=False)
self.img3.setImage(self.img_array, autoLevels=False)
self.img4.setImage(self.img_array, autoLevels=False)
self.img5.setImage(self.img_array, autoLevels=False)
self.img6.setImage(self.img_array, autoLevels=False)
self.img7.setImage(self.img_array, autoLevels=False)
self.img8.setImage(self.img_array, autoLevels=False)
Grant R
Dec 17, 2019, 3:41:10 PM12/17/19
to pyqtgraph
I think it would be much cleaner if you had a list of image items, so self.imageItems = [self.img, self.img1, ...]. Then you can set all your labels, scale, etc in one loop rather than repeating the code. You can pass the complete 'y' from read into your update method. Your img_array isn't indexed by channel but needs to be. Every ImageItem is being passed the same img_array so the plots will be the same.
def read(self):
self.stream.start_stream()
data = self.stream.read(CHUNKSZ)
y = np.fromstring(data, dtype=np.int16)
self.signal.emit(y)
...
def update(self, all_chunk):
# normalized, windowed frequencies in data chunk
for i in len(self.imageItems):
chunk = all_chunk[i::8] # Channel i
spec = np.fft.rfft(chunk*self.win2) / CHUNKSZ # Compute discrete FT for real input
#print(spec)
# get magnitude
psd = abs(spec)
# convert to dB scale
psd = 20 * np.log10(psd)
#### img_array needs to be indexed by channel. Right now it will just repeat the same data across all plots
# roll down one and replace leading edge with new data
self.img_array = np.roll(self.img_array, -1, 0)
self.img_array[-1:] = psd
self.imageItems[i].setImage(self.img_array_by_channel, autoLevels=False)
Aldrich Cabrera
Dec 19, 2019, 2:38:48 PM12/19/19
to pyqtgraph
It is working now. :) Thank you very much for the support! I appreciate your help Grant R.
Now I need to save images every 2 seconds. However the computer is slow, I suppose because I am reading all the channels and plotting at the same time.
Maybe it will be comfortable and fast in OpenCV or maybe no. What do you think?
For now, I will have to save the images. Thanks for the support!
Grant Roch
Dec 19, 2019, 3:23:36 PM12/19/19
to pyqt...@googlegroups.com
What are you having trouble doing? Plotting or plotting and saving? The plotting should be quite fast and you can do the saving similarly to how you plot. Have another qtimer and iterate through the images saving each time.
--
You received this message because you are subscribed to a topic in the Google Groups "pyqtgraph" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/pyqtgraph/_RjdXBDKRJg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to pyqtgraph+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pyqtgraph/1fde7c22-5c5c-406d-91f8-2647d1d27ce6%40googlegroups.com.
Aldrich Cabrera
Dec 19, 2019, 5:01:29 PM12/19/19
to pyqtgraph
Yes, I want to plot and saving, but when I am doing this, the computer becomes slow.
You suggest that I will add another Qtimer to save the images while plotting the channels, I will do that.
El jueves, 19 de diciembre de 2019, 14:23:36 (UTC-6), Grant R escribió:
What are you having trouble doing? Plotting or plotting and saving? The plotting should be quite fast and you can do the saving similarly to how you plot. Have another qtimer and iterate through the images saving each time.
On Fri, Dec 20, 2019 at 8:38 AM Aldrich Cabrera <alucar...@gmail.com> wrote:
--It is working now. :) Thank you very much for the support! I appreciate your help Grant R.Now I need to save images every 2 seconds. However the computer is slow, I suppose because I am reading all the channels and plotting at the same time.Maybe it will be comfortable and fast in OpenCV or maybe no. What do you think?For now, I will have to save the images. Thanks for the support!
You received this message because you are subscribed to a topic in the Google Groups "pyqtgraph" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/pyqtgraph/_RjdXBDKRJg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to pyqt...@googlegroups.com.
Aldrich Cabrera
Feb 14, 2020, 8:02:52 PM2/14/20
to pyqtgraph
HI. I'm again. I have i question about the plotting and saving images using another qtimer. I would like to save images in the same time that plotting the spectrograms. Actually, I save the images using this:
def saveFig(self):
fname1 = "/home/oyuki/Downloads/spectro_Real_time/audios/mic1/result_%s.png"%time.strftime("%H:%M:%S", gmtime())
self.exporter1 = pg.exporters.ImageExporter(self.imageItems[0])
self.exporter2 = pg.exporters.ImageExporter(self.imageItems[1])
self.exporter3 = pg.exporters.ImageExporter(self.imageItems[2])
self.exporter4 = pg.exporters.ImageExporter(self.imageItems[3])
self.exporter5 = pg.exporters.ImageExporter(self.imageItems[4])
self.exporter6 = pg.exporters.ImageExporter(self.imageItems[5])
self.exporter7 = pg.exporters.ImageExporter(self.imageItems[6])
self.exporter8 = pg.exporters.ImageExporter(self.imageItems[7])
self.exporter = [self.exporter1,self.exporter2,self.exporter3,self.exporter4,self.exporter5,self.exporter6,self.exporter7,self.exporter8]
for i in range(len(self.pItems)):
self.exporter[i].params.param('width').setValue(304, blockSignal=self.exporter[i].widthChanged)
self.exporter[i].params.param('height').setValue(163, blockSignal=self.exporter[i].heightChanged)
self.exporter[0].export(fname1)
El jueves, 19 de diciembre de 2019, 14:23:36 (UTC-6), Grant R escribió:
What are you having trouble doing? Plotting or plotting and saving? The plotting should be quite fast and you can do the saving similarly to how you plot. Have another qtimer and iterate through the images saving each time.
On Fri, Dec 20, 2019 at 8:38 AM Aldrich Cabrera <alucar...@gmail.com> wrote:
--It is working now. :) Thank you very much for the support! I appreciate your help Grant R.Now I need to save images every 2 seconds. However the computer is slow, I suppose because I am reading all the channels and plotting at the same time.Maybe it will be comfortable and fast in OpenCV or maybe no. What do you think?For now, I will have to save the images. Thanks for the support!
You received this message because you are subscribed to a topic in the Google Groups "pyqtgraph" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/pyqtgraph/_RjdXBDKRJg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to pyqt...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages