def extract(mp3file,npyfile=''):
"""
Get an mp3 file, reads it, extract fingerprints features,
saves it to a given npyfile (if different than '')
RETURN numpy array for debugging purposes
"""
# read mp3 through MATLAB mp3 file
x,fs,tmp = audioio.audioread(mp3file)
x = np.average(x,axis=1)
assert x.shape[0] > 2,'bad signal averaging'
# get the fingerprint features
L,S,T,maxes = LANDMARKS.find_landmarks(x,fs)
# create the maxes matrix
maxes = maxes - 1 # matlab starts at 0
nfreq = 256
assert np.max(maxes[1,:]) < nfreq,'max freq too high'
nsamp = int(np.max(maxes[0,:]) + 1)
data = np.zeros([nfreq,nsamp])
# fill in data
for m in range(maxes.shape[1]):
data[maxes[1,m],maxes[0,m]] += 1
# save to npy
if npyfile != '':
np.save(npyfile,data)
# return data for debugging or ipython usage
return data
def maxes_beattimes_segs_from_audiofile(wavfile):
"""
Utility function
From a given eavfile:
- computes the fingerprint maxes
- gets the corresponding en beat times
- gets the corresponding annotated segments
RETURN:
signal, sampling rate, maxes, beatstarts, duration, segstarts, labels
"""
# fingerprint it
wav = LANDMARKS.AUDIOLAB.wavread(wavfile)
L,S,T,maxes = LANDMARKS.find_landmarks(wav[0],wav[1])
# find the EN matfile
relwavfile = os.path.relpath(wavfile,start=_audio_dir)
enmatfile = os.path.join(_enfeats_dir,relwavfile+'.mat')
assert os.path.exists(enmatfile),'can not find matfile %s' % enmatfile
mat = sio.loadmat(enmatfile)
btstart = mat['btstart']
try:
duration = mat['duration'][0][0] # must be some bug in encoding
except TypeError:
duration = mat['duration']
# get the segments
labfile = enmatfile+'.lab'
assert os.path.exists(labfile),'can not find labfile %s' % labfile
segstarts, segstops, labels = DUMMY.read_lab_file(labfile)
# done, return
return wav[0], wav[1], maxes, btstart, duration, segstarts, labels
def maxes_beattimes_segs_from_audiofile(wavfile):
"""
Utility function
From a given eavfile:
- computes the fingerprint maxes
- gets the corresponding en beat times
- gets the corresponding annotated segments
RETURN:
signal, sampling rate, maxes, beatstarts, duration, segstarts, labels
"""
# fingerprint it
wav = LANDMARKS.AUDIOLAB.wavread(wavfile)
L, S, T, maxes = LANDMARKS.find_landmarks(wav[0], wav[1])
# find the EN matfile
relwavfile = os.path.relpath(wavfile, start=_audio_dir)
enmatfile = os.path.join(_enfeats_dir, relwavfile + '.mat')
assert os.path.exists(enmatfile), 'can not find matfile %s' % enmatfile
mat = sio.loadmat(enmatfile)
btstart = mat['btstart']
try:
duration = mat['duration'][0][0] # must be some bug in encoding
except TypeError:
duration = mat['duration']
# get the segments
labfile = enmatfile + '.lab'
assert os.path.exists(labfile), 'can not find labfile %s' % labfile
segstarts, segstops, labels = DUMMY.read_lab_file(labfile)
# done, return
return wav[0], wav[1], maxes, btstart, duration, segstarts, labels
