Obsolete API¶

Overview¶

The following code shows you how to open a file for read, reading the first 1000 frames, and closing it:

import scikits.audiolab as  audiolab

a       = audiolab.sndfile('test.wav', 'read')
data    = a.read_frames(1000)
a.close()

Opening a file and getting its parameters¶

Once imported, audiolab gives you access the sndfile class, which is the class of audiolab use to open audio files. You create a sndfile instance when you want to open a file for reading or writing (the file test.flac is included in the audiolab package, in the test_data directory):

import scikits.audiolab as audiolab

filename    = 'test.wav'
a           = audiolab.sndfile(filename, 'read')

print a

Prints you the information related to the file, like its sampling rate, the number of frames, etc... You can of course get each parameter individually by using the corresponding sndfile.get* accessors.

Importing audio data¶

Now that we’ve opened a file, we would like to read its audio content, right ? For now, you can only import the data as floating point data, float (32 bits) or double (64 bits). The function sndfile.read_frames read n frames, where a frame contains a sample of each channel (one in mono, 2 in stereo, etc...):

import numpy as N

import scikits.audiolab as audiolab

filename    = 'test.wav'
a           = audiolab.sndfile(filename, 'read')

tmp         = a.read_frames(1e4)
float_tmp   = a.read_frames(1e4, dtype = N.float32)

import pylab as P
P.plot(tmp[:])

The above code import 10000 frames, and plot the first channel using matplotlib (see below). A frame holds one sample from each channel: 1000 frames of a stereo file is 2000 samples. Each channel is one column of the numpy array. The read functions follow numpy conventions, that is by default, the data are read as double, but you can give a dtype argument to the function.

The format class¶

When opening a file for writing, you need to give various parameters related to the format such as the file format, the encoding. The format class is used to create valid formats from those parameters By default, the format class creates a format object with file type wav, and 16 bits pcm encoding:

from scikits.audiolab import formatinfo as format

f = format('aiff', 'ulaw')
print f

f = format('ircam', 'float32')
print f

prints back “Major Format: AIFF (Apple/SGI), Encoding Format: U-Law” and “Major Format: SF (Berkeley/IRCAM/CARL), Encoding Format: 32 bit float”.

Writing data to a file¶

Opening a file for writing is a bit more complicated than reading; you need to say which format you are requesting, the number of channels and the sampling rate (in Hz) you are requesting; all those information are mandatory ! The class format is used to build a format understandable by libsndfile from ‘user-friendly’ values. Let’s see how it works.

from tempfile import mkstemp
from os import remove

import numpy as N
from scikits.audiolab import formatinfo as format
import scikits.audiolab as audiolab

# Create a temp file in the system temporary dir, and always remove
# it at the end
cd, filename    = mkstemp('tmptest.wav')
try:
    fmt         = format('wav', 'pcm24')
    nchannels   = 2
    fs          = 44100

    afile =  audiolab.sndfile(filename, 'write', fmt, nchannels, fs)

    # Create a stereo white noise, with Gaussian distribution
    tmp = 0.1 * N.random.randn(1000, nchannels)

    # Write the first 500 frames of the signal
    # Note that the write_frames method uses tmp's numpy dtype to determine how
    # to write to the file; sndfile also converts the data on the fly if necessary
    afile.write_frames(tmp, 500)

    afile.close()

    # Let's check that the written file has the expected meta data
    afile = audiolab.sndfile(filename, 'read')
    assert(afile.get_samplerate() == fs)
    assert(afile.get_channels() == nchannels)
    assert(afile.get_nframes() == 500)
finally:
    remove(filename)