def audio_features(params, img_audio, audio_path, append_name, node_list):
output_file = params[5]
# create pytable atom for the features
f_atom = tables.Float32Atom()
count = 1
# keep track of the nodes for which no features could be made, places database contains some
# empty audio files
invalid = []
for node in node_list:
print('processing file audio:' + str(count))
count += 1
# create a group for the desired feature type (e.g. a group called 'fbanks')
audio_node = output_file.create_group(node, params[4])
# get the base name of the node this feature will be appended to
base_name = node._v_name.split(append_name)[1]
# get the caption file names corresponding to the image of this node
caption_files = [base_name + '.wav']
for cap in caption_files:
# basename for the caption file, i.e. cut of the file extension as dots arent
# allowed in pytables group names.
base_capt = cap.split('.')[0]
# as the places database splits the audio files over multiple subfolders these paths from
# the top folder are included the captions in the dictionary but can be removed from the base_name
# of the node in the h5 file.
if '/' in base_capt:
base_capt = base_capt.split('/')[-1]
# read audio samples
try:
input_data = read(os.path.join(audio_path, cap))
# in the places database some of the audiofiles are empty. To keep this script
# compatible with database that might have more captions to one image, we check
# if the audio node is empty at the end of the loop and delete the entire node
# if no caption features could be made.
if len(input_data[1]) == 0:
# break as we can do nothing with an empty audio file.
break
except:
# try to repair the file, however I found some files in places, so broken that
# such that they could not be read at all. Just remove such nodes
try:
fix_wav(os.path.join(audio_path, cap))
input_data = read(os.path.join(audio_path, cap))
except:
break
# sampling frequency
fs = input_data[0]
# get window and frameshift size in samples
window_size = int(fs * params[2])
frame_shift = int(fs * params[3])
# create features (implemented are raw audio, the frequency spectrum, fbanks and
# mfcc's)
if params[4] == 'raw':
[features, energy] = raw_frames(input_data, frame_shift, window_size)
elif params[4] == 'freq_spectrum':
[frames, energy] = raw_frames(input_data, frame_shift, window_size)
features = get_freqspectrum(frames, params[0], fs, window_size)
elif params[4] == 'fbanks':
[frames, energy] = raw_frames(input_data, frame_shift, window_size)
freq_spectrum = get_freqspectrum(frames, params[0], fs, window_size)
features = get_fbanks(freq_spectrum, params[1], fs)
elif params[4] == 'mfcc':
[frames, energy] = raw_frames(input_data, frame_shift, window_size)
freq_spectrum = get_freqspectrum(frames, params[0], fs, window_size)
fbanks = get_fbanks(freq_spectrum, params[1], fs)
features = get_mfcc(fbanks)
# optionally add the frame energy
if params[7]:
features = numpy.concatenate([energy[:, None], features], 1)
# optionally add the deltas and double deltas
if params[6]:
single_delta = delta(features, 2)
double_delta = delta(single_delta, 2)
features = numpy.concatenate([features, single_delta, double_delta], 1)
# create new leaf node in the feature node for the current audio file
feature_shape = numpy.shape(features)[1]
f_table = output_file.create_earray(audio_node, append_name + base_capt, f_atom, (0, feature_shape),
expectedrows=5000)
# append new data to the tables
f_table.append(features)
if audio_node._f_list_nodes() == []:
# keep track of all the invalid nodes for which no features could be made
invalid.append(node._v_name)
# remove the top node including all other features if no captions features could be created
output_file.remove_node(node, recursive=True)
print(invalid)
return
def audio_features (params, audio_path, append_name, node_list):
output_file = params[5]
# create pytable atom for the features
f_atom= tables.Float32Atom()
count = 1
# keep track of the nodes for which no features could be made, basically empty audio files
invalid = []
for node in node_list:
print('processing file:' + str(count))
count+=1
# create a group for the desired feature type (e.g. a group called 'fbanks')
audio_node = output_file.create_group(node, params[4])
# get the base name of the node this feature will be appended to
base_name = node._v_name.split(append_name)[1]
lis = base_name.split("_")
participant = lis[0]
emotion = lis[1]
word = lis[2]
audio_file = "_".join([participant,word,emotion]) + ".wav"
audio_folder = "_".join([participant,emotion])
input_data = read(os.path.join(audio_path, audio_folder + "\\" + audio_file))
if len(input_data[1]) == 0:
print("$")
# sampling frequency
fs = input_data[0]
# get window and frameshift size in samples
window_size = int(fs*params[2])
frame_shift = int(fs*params[3])
# create features (implemented are raw audio, the frequency spectrum, fbanks and
# mfcc's)
if params[4] == 'raw':
[features, energy] = raw_frames(input_data, frame_shift, window_size)
elif params[4] == 'freq_spectrum':
[frames, energy] = raw_frames(input_data, frame_shift, window_size)
features = get_freqspectrum(frames, params[0], fs, window_size)
elif params[4] == 'fbanks':
[frames, energy] = raw_frames(input_data, frame_shift, window_size)
freq_spectrum = get_freqspectrum(frames, params[0], fs, window_size)
features = get_fbanks(freq_spectrum, params[1], fs)
elif params[4] == 'mfcc':
[frames, energy] = raw_frames(input_data, frame_shift, window_size)
freq_spectrum = get_freqspectrum(frames, params[0], fs, window_size)
fbanks = get_fbanks(freq_spectrum, params[1], fs)
features = get_mfcc(fbanks)
# optionally add the frame energy
if params[7]:
features = numpy.concatenate([energy[:,None], features],1)
# optionally add the deltas and double deltas
if params[6]:
single_delta= delta (features,2)
double_delta= delta(single_delta,2)
features= numpy.concatenate([features,single_delta,double_delta],1)
# create new leaf node in the feature node for the current audio file
feature_shape= numpy.shape(features)[1] #39
f_table = output_file.create_earray(audio_node, append_name + base_name, f_atom, (0,feature_shape),expectedrows=5000)
# append new data to the tables
f_table.append(features)
if audio_node._f_list_nodes() == []:
# keep track of all the invalid nodes for which no features could be made
invalid.append(node._v_name)
# remove the top node including all other features if no captions features could be created
output_file.remove_node(node, recursive = True)
print(invalid)
return
assert (fs_target == fs)
# set the fft size to the power of two equal to or greater than
# the window size.
window_size = int(fs * params['t_window'])
exp = 1
while True:
if np.power(2, exp) - window_size >= 0:
fft_size = np.power(2, exp)
break
else:
exp += 1
window_size = int(fs * params['t_window'])
frame_shift = int(fs * params['t_shift'])
[frames, energy] = raw_frames(prime_data, frame_shift, window_size)
freq_spectrum = get_freqspectrum(frames, params['alpha'], fs, window_size)
fbanks = get_fbanks(freq_spectrum, params['n_filters'], fs)
prime_features = get_mfcc(fbanks)
prime_features = np.concatenate([energy[:, None], prime_features], 1)
[frames, energy] = raw_frames(target_data, frame_shift, window_size)
freq_spectrum = get_freqspectrum(frames, params['alpha'], fs, window_size)
fbanks = get_fbanks(freq_spectrum, params['n_filters'], fs)
target_features = get_mfcc(fbanks)
target_features = np.concatenate([energy[:, None], target_features], 1)
single_delta = base.delta(prime_features, params['delta_n'])
double_delta = base.delta(single_delta, params['delta_n'])