Using Python to plot the current microphone's input and the Fourier Transform

streamAudio.py

try:
    import pyaudio
    import numpy as np
    import pylab
    import matplotlib.pyplot as plt
    from scipy.io import wavfile
    import time
    import sys
    import seaborn as sns
except:
    print "Something didn't import"

i=0
f,ax = plt.subplots(2)

# Prepare the Plotting Environment with random starting values
x = np.arange(10000)
y = np.random.randn(10000)

# Plot 0 is for raw audio data
li, = ax[0].plot(x, y)
ax[0].set_xlim(0,1000)
ax[0].set_ylim(-5000,5000)
ax[0].set_title("Raw Audio Signal")
# Plot 1 is for the FFT of the audio
li2, = ax[1].plot(x, y)
ax[1].set_xlim(0,5000)
ax[1].set_ylim(-100,100)
ax[1].set_title("Fast Fourier Transform")
# Show the plot, but without blocking updates
plt.pause(0.01)
plt.tight_layout()

FORMAT = pyaudio.paInt16 # We use 16bit format per sample
CHANNELS = 1
RATE = 44100
CHUNK = 1024 # 1024bytes of data red from a buffer
RECORD_SECONDS = 0.1
WAVE_OUTPUT_FILENAME = "file.wav"

audio = pyaudio.PyAudio()

# start Recording
stream = audio.open(format=FORMAT,
                    channels=CHANNELS,
                    rate=RATE,
                    input=True)#,
                    #frames_per_buffer=CHUNK)

global keep_going
keep_going = True

def plot_data(in_data):
    # get and convert the data to float
    audio_data = np.fromstring(in_data, np.int16)
    # Fast Fourier Transform, 10*log10(abs) is to scale it to dB
    # and make sure it's not imaginary
    dfft = 10.*np.log10(abs(np.fft.rfft(audio_data)))

    # Force the new data into the plot, but without redrawing axes.
    # If uses plt.draw(), axes are re-drawn every time
    #print audio_data[0:10]
    #print dfft[0:10]
    #print
    li.set_xdata(np.arange(len(audio_data)))
    li.set_ydata(audio_data)
    li2.set_xdata(np.arange(len(dfft))*10.)
    li2.set_ydata(dfft)

    # Show the updated plot, but without blocking
    plt.pause(0.01)
    if keep_going:
        return True
    else:
        return False

# Open the connection and start streaming the data
stream.start_stream()
print "\n+---------------------------------+"
print "| Press Ctrl+C to Break Recording |"
print "+---------------------------------+\n"

# Loop so program doesn't end while the stream callback's
# itself for new data
while keep_going:
    try:
        plot_data(stream.read(CHUNK))
    except KeyboardInterrupt:
        keep_going=False
    except:
        pass

# Close up shop (currently not used because KeyboardInterrupt
# is the only way to close)
stream.stop_stream()
stream.close()

audio.terminate()

Hey, this code is very useful except for when each time I try to run it, first at all, it did show the figure,
but later my computer juste shut down the figure, I tried to change de Record sounds but it doesn't change anything.
Did anyone have de same problem as I ?

Hey, this code is very useful except for when each time I try to run it, first at all, it did show the figure,
but later my computer juste shut down the figure, I tried to change de Record sounds but it doesn't change anything.
Did anyone have de same problem as I ?

I had the same problem as you. It would show two frames of the FFT and then freeze.

If you remove the try catch block at the bottom, you see that this code raises an "Input Overflow" pyaudio Exception. If I pass an argument to stream.read called exception_on_overflow set to False (and add parentheses to all of the print statements), then this code works for me.

One issue that I keep encountering is that every time that I record and plot real-time audio input signals I get some small repetitive gaps in the recording due to the plotting delay. If you simply store the streaming audio in the above code and then store is as a wave file you'll notice small gaps, which will disappear once you stop plotting the data. Any ideas on how to solve this? Multi-processing maybe?

Hi

I'm trying to run this code but it freezes just after beginning, it seems to work, though, but just for a fraction of a second.

I tried to follow instructions on landa's comment but I don't know how to pass the argument to stream.read, I managed to remove the try block, and now it gives me the input overflow exception.

Any tips on this?

Thanks in advance

I’m on my phone right how, but you pass this argument like any other argument in Python: function(arg1, arg2, named_argument=value).

…

Sent from my iPhone

On Mar 10, 2020, at 2:53 PM, Vnaf ***@***.***> wrote:  Hi I'm trying to run this code but it freezes just after beginning, it seems to work, though, but just for a fraction of a second. I tried to follow instructions on landa's comment but I don't know how to pass the argument to stream.read, I managed to remove the try block, and now it gives me the input overflow exception. Any tips on this? Thanks in advance — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

Yes, I just found it, thanks!

Line 87 plot_data(stream.read(CHUNK)) -->plot_data(stream.read(CHUNK, exception_on_overflow = False))

Hey, this code is very useful except for when each time I try to run it, first at all, it did show the figure, but later my computer juste shut down the figure, I tried to change de Record sounds but it doesn't change anything. Did anyone have de same problem as I ?

Maybe try running the code without the "try:" block. As that block is used to check for errors in code, maybe not having any errors in the first place is causing the code to end.

line67 li2.set_xdata(np.arange(len(dfft))*10.)
li2.set_xdata(np.arange(len(dfft))*RATE/CHUNK)
frequency should be scaled by a factor dependent on the sampling rate and chunk
right now its reporting a 440hz signal as ~110hz