Pop-up window of Audio-spectrum analyzer is appearing but it seems that there is no signal input as there is a non moving flat line on the 2 displays

[bjce]

Hello Mark

First of all all thank you very much for the very nice video on youtube with great ideas and the script that you provide on your website. Those are very interesting projects!

I'm trying in macOS 10.14 and run then code of Spectrum analyzer that you provide with the following code.

"""
Notebook for streaming data from a microphone in realtime

audio is captured using pyaudio
then converted from binary data to ints using struct
then displayed using matplotlib

scipy.fftpack computes the FFT

if you don't have pyaudio, then run

>>> pip install pyaudio

note: with 2048 samples per chunk, I'm getting 20FPS
when also running the spectrum, its about 15FPS
"""

import pyaudio
import os
import struct
import numpy as np
import matplotlib.pyplot as plt
from scipy.fftpack import fft
import time
from tkinter import TclError

# to display in separate Tk window
%matplotlib tk

# constants
CHUNK = 1024 * 2             # samples per frame
FORMAT = pyaudio.paInt16     # audio format (bytes per sample?)
CHANNELS = 1                 # single channel for microphone
RATE = 44100                 # samples per second

# create matplotlib figure and axes
fig, (ax1, ax2) = plt.subplots(2, figsize=(15, 7))

# pyaudio class instance
p = pyaudio.PyAudio()

# stream object to get data from microphone
stream = p.open(
    format=FORMAT,
    channels=CHANNELS,
    rate=RATE,
    input=True,
    output=True,
    frames_per_buffer=CHUNK
)

# variable for plotting
x = np.arange(0, 2 * CHUNK, 2)       # samples (waveform)
xf = np.linspace(0, RATE, CHUNK)     # frequencies (spectrum)

# create a line object with random data
line, = ax1.plot(x, np.random.rand(CHUNK), '-', lw=2)

# create semilogx line for spectrum
line_fft, = ax2.semilogx(xf, np.random.rand(CHUNK), '-', lw=2)

# format waveform axes
ax1.set_title('AUDIO WAVEFORM')
ax1.set_xlabel('samples')
ax1.set_ylabel('volume')
ax1.set_ylim(0, 255)
ax1.set_xlim(0, 2 * CHUNK)
plt.setp(ax1, xticks=[0, CHUNK, 2 * CHUNK], yticks=[0, 128, 255])

# format spectrum axes
ax2.set_xlim(20, RATE / 2)

print('stream started')

# for measuring frame rate
frame_count = 0
start_time = time.time()

while True:

    # binary data
    data = stream.read(CHUNK)

    # convert data to integers, make np array, then offset it by 127
    data_int = struct.unpack(str(2 * CHUNK) + 'B', data)

    # create np array and offset by 128
    data_np = np.array(data_int, dtype='b')[::2] + 128

    line.set_ydata(data_np)

    # compute FFT and update line
    yf = fft(data_int)
    line_fft.set_ydata(np.abs(yf[0:CHUNK])  / (128 * CHUNK))

    # update figure canvas
    try:
        fig.canvas.draw()
        fig.canvas.flush_events()
        frame_count += 1

    except TclError:

        # calculate average frame rate
        frame_rate = frame_count / (time.time() - start_time)

        print('stream stopped')
        print('average frame rate = {:.0f} FPS'.format(frame_rate))
        break

I made sure to allow Atom (my text editor) and nteract to the microphone.

There is a pop-up window for both spectrum and wave appearance. However the lines of both those lines stays flat as follow as if there was no signal input:

At the same time I got the following error message in the text editor:

stream started
---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
<ipython-input-4-7818c8173441> in <module>()
     80 
     81     # binary data
---> 82     data = stream.read(CHUNK)
     83 
     84     # convert data to integers, make np array, then offset it by 127

/anaconda3/lib/python3.6/site-packages/pyaudio.py in read(self, num_frames, exception_on_overflow)
    606                           paCanNotReadFromAnOutputOnlyStream)
    607 
--> 608         return pa.read_stream(self._stream, num_frames, exception_on_overflow)
    609 
    610     def get_read_available(self):

OSError: [Errno -9981] Input overflowed

Do you have any idea of where the problem could be

[rudrathegreat]

Probably because there is no data coming through the stream. You need to define what the recording device is, that is where the data will be coming from.

See, pyaudio's open function simply turns on the recording device, so you need to define the recording device. I had similar problems as well and this is how I solved them

[rudrathegreat]

Similar to this issue - #5
Duplicate of #5