def audio(filename, name_file, hpss, sr):
"""
Wrapper of get audio for save all results for future same calculations. If parameterization
have been computed before get audio is not computed.
Parameters
----------
filename: str
name of the file that is being computed witout format extension
name_file: str
name of the file that is being computed
hpss: bool optional
true or false depends is harmonic percussive source separation (hpss) block wants to be computed. Default False.
sr: number > 0 [scalar]
target sampling rate
Returns
-------
list
sample of audio
"""
name_audio = get_name_audio(name_file, hpss, sr)
if path.exists(name_audio):
dic = load_binary(name_audio)
else:
y, sr = get_audio(filename, hpss, sr)
dic = {'y': y, 'sr': sr}
# dic_save = {'y': y.tolist(), 'sr': sr}
save_binary(dic, name_audio)
# save_json(dic_save, name_audio + '.json')
return dic['y'], dic['sr']
def harmonic_change(filename: str, name_file: str, hpss: bool = False, tonal_model: str = 'TIV2', chroma: str = 'cqt',
blur: str = 'full', sigma: int = 11, log_compresion: str = 'none', distance: str = 'euclidean'):
"""
Wrapper of harmonic change detection function for save all results for future same calculations. If parameterization
have been computed before HCDF is not computed.
Parameters
----------
filename: str
name of the file that is being computed witout format extension
name_file: str
name of the file that is being computed
hpss : bool optional
true or false depends is harmonic percussive source separation (hpss) block wants to be computed. Default False.
tonal_model: str optional
Tonal model block type. "TIV2" for Tonal Interval space focus on audio. "TIV2" for audio. "TIV2_Symb" for symbolic data.
"tonnetz" for harte centroids aproach. Default TIV2
chroma: str optional
"chroma-samplerate-framesize-overlap"
chroma can be "CQT","NNLS", "STFT", "CENS" or "HPCP"
samplerate as a number scalar
frame size as a number scalar
overlap number that a windows is divided
sigma: number (scalar > 0) optional
sigma of gaussian smoothing value. Default 11
distance: str optional
type of distance measure used. Types can be "euclidean" for euclidean distance and "cosine" for cosine distance. Default "euclidean".
Returns
-------
list
harmonic changes (the peaks) on the song detected
list
HCDF function values
number
windows size
"""
centroid_changes = []
check_parameters(chroma, blur, tonal_model, log_compresion, distance)
name_harmonic_change = get_name_harmonic_change(name_file, hpss, tonal_model, chroma, blur, sigma, log_compresion,
distance)
if path.exists(name_harmonic_change):
dic = load_binary(name_harmonic_change)
else:
changes, harmonic_function, windows_size, centroid_changes = get_harmonic_change(filename, name_file, hpss,
tonal_model, chroma,
blur, sigma, log_compresion,
distance)
dic = {'changes': changes, 'harmonic_function': harmonic_function, 'windows_size': windows_size}
save_binary(dic, name_harmonic_change)
return dic['changes'], dic['harmonic_function'], dic['windows_size']
def gaussian_blur(hpss, chroma, tonal_model, name_file, centroid_vector,
log_compresion, blur, sigma):
"""
Wrapper of get_gaussian_blur for save all results for future same calculations. If parameterization
have been computed before get_gaussian_blur is not computed.
Parameters
----------
name_file: str
name of the file that is being computed
hpss: bool optional
true or false depends is harmonic percussive source separation (hpss) block wants to be computed. Default False.
sr: number > 0 [scalar]
target sampling rate
chroma: str optional
"chroma-samplerate-framesize-overlap"
chroma can be "CQT","NNLS", "STFT", "CENS" or "HPCP"
samplerate as a number scalar
frame size as a number scalar
overlap number that a windows is divided
tonal_model: str optional
Tonal model block type. "TIV2" for Tonal Interval space focus on audio. "TIV2" for audio. "TIV2_Symb" for symbolic data.
"tonnetz" for harte centroids aproach. Default TIV2
centoid_vector: list
tonal centroids of the tonal model
sigma: number (scalar > 0) optional
sigma of gaussian smoothing value. Default 11
Returns
-------
list
sample of audio
"""
gaussian_blur = get_name_gaussian_blur(name_file, hpss, chroma,
tonal_model, blur, sigma,
log_compresion)
if path.exists(gaussian_blur):
dic = load_binary(gaussian_blur)
else:
centroid_vector = get_gaussian_blur(centroid_vector, blur, sigma)
dic = {'centroid_vector': centroid_vector}
# dic_save = {'centroid_vector': centroid_vector.tolist()}
save_binary(dic, gaussian_blur)
# save_json(dic_save, gaussian_blur + '.json')
return dic['centroid_vector']
def tonal_centroid_transform(hpss, chroma, name_file, y, sr, tonal_model,
doce_bins_tuned_chroma):
"""
wrapper of tonal centroid transform for save all results for future same calculations
Parameters
----------
hpss : bool
true or false depends on hpss block
name_file: str
name of the file that is being computed
y : number > 0 [scalar]
audio
sr: number > 0 [scalar]
target sampling rate
chroma: str
chroma-samplerate-framesize-overlap
tonal_model: str optional
Tonal model block type. "TIV2" for Tonal Interval space focus on audio. "TIV2" for audio. "TIV2_Symb" for symbolic data.
"tonnetz" for harte centroids aproach. Default TIV2\
doce_bins_tuned_chroma: list
list of chroma vectors
Returns
-------
list of tonal centroids vectors
"""
name_tonal_model = get_name_tonal_model(name_file, hpss, chroma,
tonal_model)
if tonal_model == 'without_tc':
dic = {'centroid_vector': doce_bins_tuned_chroma}
else:
if path.exists(name_tonal_model):
dic = load_binary(name_tonal_model)
else:
centroid_vector = get_tonal_centroid_transform(
y, sr, tonal_model, doce_bins_tuned_chroma)
dic = {'centroid_vector': centroid_vector}
save_binary(dic, name_tonal_model)
return dic['centroid_vector']
def chromagram(hpss, name_file, y, sr, chroma):
"""
wrapper of get_chromagram for save all results for future same calculations
Parameters
----------
hpss : bool
true or false depends on hpss block
name_file: str
name of the file that is being computed
y : number > 0 [scalar]
audio
sr: number > 0 [scalar]
target sampling rate
chroma: str
chroma-samplerate-framesize-overlap
Returns
-------
list of chromagrams
"""
name_chromagram = get_name_chromagram(name_file, hpss, chroma)
if path.exists(name_chromagram):
dic = load_binary(name_chromagram)
else:
# if mutex_global.mutex is not None:
# mutex_global.mutex.acquire()
doce_bins_tuned_chroma = get_chromagram(y, sr, chroma)
# if mutex_global.mutex is not None:
# mutex_global.mutex.release()
dic = {'doce_bins_tuned_chroma': doce_bins_tuned_chroma}
# dic_save = {'doce_bins_tuned_chroma': doce_bins_tuned_chroma.tolist()}
save_binary(dic, name_chromagram)
# save_json(dic_save, name_chromagram + '.json')
return dic['doce_bins_tuned_chroma']