How to plot pyaudio input with matplotlib?

Question

How can I plot on matplotlib input signal from microphone? I have tried to plot with plt.plot(frames) but frames is for some reason a string?

a) Why is frames variable a string list?

b) Why is data variable string list?

c) Should they represent energy/amplitude of single sample and be integers?

d) Why is length of data 2048 when I specified I want chunk size of 1024?

(I guess because i use paInt16, but cannot see still why it couldn't be 1024)

I have the following code for microphone input:

import pyaudio
import audioop
import matplotlib.pyplot as plt
import numpy as np
from itertools import izip
import wave


FORMAT = pyaudio.paInt16                # We use 16bit format per sample
CHANNELS = 1
RATE = 44100
CHUNK = 1024                            # 1024bytes of data red from a buffer
RECORD_SECONDS = 3
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)

frames = []
for i in range(0, int(RATE / CHUNK * RECORD_SECONDS)):
    data = stream.read(CHUNK)
    frames.append(data)
frames = ''.join(frames)

stream.stop_stream()
stream.close()
audio.terminate()

Can you include your import statements with your example code? Also, the frames variable is a string list because you declare it as so, frames = ''.join(frames). You don't need to do that, since you already appended all the frames you need, you have a list. — pretzlstyle
– pretzlstyle, Commented Dec 16, 2015 at 1:04
Yes, I am aware of that but data is also a string. I think there is something with struct.unpack that needs to be done, but I have no idea what exactly. — Dusan J.
– Dusan J., Commented Dec 16, 2015 at 1:08
But you asked why frames was a list of strings? Anyway, Stream.read() is suppossed to return a string, as specified in the API Documentation: people.csail.mit.edu/hubert/pyaudio/docs — pretzlstyle
– pretzlstyle, Commented Dec 16, 2015 at 1:15
In python3 you need to put a b in front: b"".join(frames) — Hallgeir Wilhelmsen
– Hallgeir Wilhelmsen, Commented Nov 28, 2019 at 7:02

fjarri · Accepted Answer · 2015-12-16 01:14:19Z

5

a) Why is frames variable a string list?

As a consequence of b), that's how you build it in your code.

b) Why is data variable string list?

It is a byte string, that is just a raw sequence of bytes. That's what read() returns.

c) Should they represent energy/amplitude of single sample and be integers?

They are. They're just packed in a byte sequence and not in Python integers.

d) Why is length of data 2048 when I specified I want chunk size of 1024?

1024 is the number of frames. Each frame is 2 bytes long, so you get 2048 bytes.

How can I plot on matplotlib input signal from microphone? I have tried to plot with plt.plot(frames) but frames is for some reason a string?

Depends on what you want to plot. Just raw amplitude can be obtained by transforming the byte string to a numpy array:

fig = plt.figure()
s = fig.add_subplot(111)
amplitude = numpy.fromstring(frames, numpy.int16)
s.plot(amplitude)
fig.savefig('t.png')

A more useful plot would be a spectrogram:

fig = plt.figure()
s = fig.add_subplot(111)
amplitude = numpy.fromstring(frames, numpy.int16)
s.specgram(amplitude)
fig.savefig('t.png')

But you can tinker with amplitude however you want, now that you have a numpy array.

answered Dec 16, 2015 at 1:14

fjarri

9,74645 silver badges52 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

pretzlstyle Over a year ago

why did you have to declare the numpy array with dtype=int16?

fjarri Over a year ago

numpy can't deduce from a raw byte string what kind of variables are stored there, you have to specify it explicitly. Since the data is read as pyaudio.paInt16, the string contains 2-byte signed integers, and that's what I tell numpy to read.

pretzlstyle Over a year ago

I see. I am not OP and I know nothing about audio, but is 16bit a standard sample size, meaning the measured intensity values have 16-bits of percision? And, is a "sample" simply one "unit" of sound, the duration of which produces one discrete intensity value?

fjarri Over a year ago

Kind of standard, I guess, for non-professionals at least. I never used pyaudio myself, but I suspect it would convert the resolution to 16 bit if it wasn't 16-bit in the original file. Yes, a frame is a single measurement (taken 44100 times per second). It may contain several channels (stereo etc), but there's only one in this case.

Collectives™ on Stack Overflow

How to plot pyaudio input with matplotlib?

1 Answer 1

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related