NumPy array built from audio file throws zero-size array error when array is neither empty nor corrupted

Question

My code is:

import numpy
from scipy.io.wavfile import read

audio_file_location = 'file_location'

audio_file = read(audio_file_location)
n = numpy.array( audio_file[1],dtype=float )
size = n.size
w = 410
limit = 205
delta  = n.size/410
i = 0
j = 0
a = 1

while i < w:
    J = min(size, j+delta)
    if J > j:
        b = numpy.max( n[j:J] )
    else:
        break

    # Do stuff
    i = i+1
    j = j+delta
    if i >= limit:
        # just something to stop the script.
        # add a break point in the next line of code.
        # and then run the top part of the script two or 3 times.
        a = 1+a

When ran, it throws:

ValueError: zero-size array to reduction operation maximum which has no identity

Other NumPy arrays such as:

numpy.array(range(0, 176942))

and other audio files work ok. The error is thrown with several audio files, not just one.

What can the source of this error be?

I don't think the file is corrupted. But if it is,

What can I do to check/fix it?

Note:

The error will happen when i > limit for limit = 205, 2 or 3 iterations after this limit is reached. The values of j and J can vary depending on the length of the audio file. Examples of j and J values are: 197760, 198720; and 88786, 89217.

Full trace back:

File "<ipython-input-14-7d842d790860>", line 1, in <module>
runfile('.../Documents/wav_analizer/questoin_on_numpy.size.py', wdir='.../Documents/wav_analizer')

File ".../.local/lib/python2.7/site-packages/spyder/utils/site/sitecustomize.py", line 705, in runfile
execfile(filename, namespace)

File ".../.local/lib/python2.7/site-packages/spyder/utils/site/sitecustomize.py", line 94, in execfile
builtins.execfile(filename, *where)

File ".../Documents/wav_analizer/questoin_on_numpy.size.py", line 30, in <module>
b = numpy.max( n[j:J] )

File ".../.local/lib/python2.7/site-packages/numpy/core/fromnumeric.py", line 2320, in amax
out=out, **kwargs)

File ".../.local/lib/python2.7/site-packages/numpy/core/_methods.py", line 26, in _amax
return umr_maximum(a, axis, None, out, keepdims)

ValueError: zero-size array to reduction operation maximum which has no identity

Partial solution: ok, I partially solved the problem:

N = [[223,32],[34,653],[-123,54],[-34,23] ...[12,43],[,123,32]],

so that explains the error. However, When I try to parse N, I get an index problem:

import os
import numpy
from scipy.io.wavfile import read

working_dir = os.path.dirname(os.path.realpath(__file__))
audio_file_location = os.path.join(working_dir, 'test_sounds/guitarup_full.wav')

audio_file = read(audio_file_location)

delta_t = audio_file[0] 
n = audio_file[1]     # this is a numpy array
print "Size",  n.size
print "Shape", n.shape

for i in range(0, n.size-1):
    a = n[i][0]
    b = n[i][1]

for i in range(0, n.shape[0]-1):
    a = n[i][0]
    b = n[i][1]

Output:

"Size" 176942
"Shape" (88471, 2)

/.local/lib/python2.7/site-packages/scipy/io/wavfile.py:273: WavFileWarning: Chunk (non-data) not understood, skipping it.
WavFileWarning)

Traceback (most recent call last):

File "<ipython-input-51-7d842d790860>", line 1, in <module>
runfile('/Documents/wav_analizer/questoin_on_numpy.size.py', wdir='/Documents/wav_analizer')

File "/.local/lib/python2.7/site-packages/spyder/utils/site/sitecustomize.py", line 705, in runfile
execfile(filename, namespace)

File "/.local/lib/python2.7/site-packages/spyder/utils/site/sitecustomize.py", line 94, in execfile
builtins.execfile(filename, *where)

File "/Documents/wav_analizer/questoin_on_numpy.size.py", line 26, in <module>
a = n[i][0]

IndexError: index 88471 is out of bounds for axis 0 with size 88471

Why the discrepancy between size and shape?

Please include the full traceback. Currently we don't know what line of code throws that error. — roganjosh
– roganjosh, Commented Jul 5, 2018 at 19:10
Also, please give us enough information about the input to diagnose the problem. What's the size or shape of the array that fails? What's the value of J and j when it fails? — abarnert
– abarnert, Commented Jul 5, 2018 at 19:13
A print-debugging suggestion: add the line print(n.shape, j, J) right before b = numpy.max( n[j:J] ). — Warren Weckesser
– Warren Weckesser, Commented Jul 5, 2018 at 19:19
the error will happen when i > limit. The values of j and J can vary depending on the length of the audio file. examples of j and J values are: 197760, 198720 — mm_
– mm_, Commented Jul 5, 2018 at 19:56

J. Martinot-Lagarde · Accepted Answer · 2018-07-05 20:08:30Z

1

I'll guess that this fails for sound files with only one channel, i.e. mono instead of stereo. In this case when you create the n array you take the first element of the channel instead of the fist channel as a whole.

You can try something like:

if audio_file.ndims > 1:
    audio_file = audio_file[1]
n = numpy.array(audio_file, dtype=float)

If the data is already one dimensional, use it directly. If the data has at least 2 channels, take the second one as you did in your code.

answered Jul 5, 2018 at 20:08

J. Martinot-Lagarde

3,5902 gold badges18 silver badges17 bronze badges

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

4 Comments

mm_ Over a year ago

but I already to: n = numpy.array( audio_file[1],dtype=float ),what is the difference?

mm_ Over a year ago

I get: AttributeError: 'tuple' object has no attribute 'ndims'

mm_ Over a year ago

The above is not a solution. See: docs.scipy.org/doc/scipy-0.19.0/reference/generated/… scipy.io.wavfile.read returns (int,numpy.array)

J. Martinot-Lagarde Over a year ago

My bad, I thought the function returned only the data. I suppose that the origin of the error is still the number of channels, but my fix is wrong.

mm_ · Accepted Answer · 2018-07-08 17:02:29Z

0

Solution:

I found a solution. The following script works:

import os
import scipy.io.wavfile
import numpy

audio_file_location = "your audio file location"

F = scipy.io.wavfile.read(audio_file_location)

delta_t = F[0]
N = F[1]

print "shape:", N.shape
print "dtype:", N.dtype
print "min, max:", N.min(), N.max()
print "first value:", N[0]
print ""

chennel_1 = numpy.zeros(N.size)
chennel_2 = numpy.zeros(N.size)

i = 0
for x in N:
    chennel_1[i] = x[0]
    chennel_2[i] = x[1]
    i = i+1

i = 0
while i < 100:
    print "ch_1", i, chennel_1[i]
    print "ch_2", i, chennel_2[i]
    i = i+1

Still don't understand the discrepancy in the shape vs size. If anyone can explain, that would be great.

answered Jul 8, 2018 at 17:02

mm_

1,7352 gold badges20 silver badges42 bronze badges

Collectives™ on Stack Overflow

NumPy array built from audio file throws zero-size array error when array is neither empty nor corrupted

2 Answers 2

4 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

4 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related