def update_read_command(self):
"""
update mic
"""
# read chunk
is_onset = self.read_mic_data()
# onset was detected
if is_onset:
# start collection of items
self.collector.start_collecting()
# collection is full
if self.collector.is_full():
# read out collection
x_onset = self.collector.read_collection()
# extract features
mfcc_bon, bon_pos = self.feature_extractor.extract_mfcc(x_onset)
# classify collection
y_hat, label = self.classifier.classify(mfcc_bon)
# plot
plot_mfcc_profile(
x_onset[bon_pos *
self.hop:(bon_pos +
self.feature_params['frame_size']) *
self.hop],
self.feature_params['fs'],
self.N,
self.hop,
mfcc_bon,
frame_size=self.feature_params['frame_size'],
plot_path=self.plot_path,
name='collect-{}_label-{}'.format(
self.collector.collection_counter, label),
enable_plot=self.mic_params['enable_plot'])
# clear read queue
self.clear_mic_queue()
return label
return None
def showcase_wavs(cfg,
raw_plot=True,
spec_plot=True,
mfcc_plot=True,
show_plot=False):
"""
showcase wavs
"""
# plot path
plot_path = '../docu/thesis/3_signal/figs/'
# change params
feature_params = cfg['feature_params'].copy()
feature_params['n_ceps_coeff'] = 32
feature_params['norm_features'] = True
# init feature extractor
feature_extractor = FeatureExtractor(feature_params)
# wav, anno dir
wav_dir, anno_dir = '../ignore/my_recordings/showcase_wavs/', '../ignore/my_recordings/showcase_wavs/annotation/'
# analyze some wavs
for wav, anno in zip(glob(wav_dir + '*.wav'),
glob(anno_dir + '*.TextGrid')):
# info
print("\nwav: ", wav), print("anno: ", anno)
# load file
x, _ = librosa.load(wav, sr=feature_params['fs'])
# raw waveform
if raw_plot:
plot_waveform(x,
feature_params['fs'],
anno_file=anno,
hop=feature_extractor.hop,
plot_path=plot_path,
name='signal_raw_' +
wav.split('/')[-1].split('.')[0] + '_my',
show_plot=show_plot)
# spectogram
if spec_plot:
plot_spec_profile(x,
feature_extractor.calc_spectogram(x).T,
feature_params['fs'],
feature_extractor.N,
feature_extractor.hop,
anno_file=anno,
plot_path=plot_path,
title=wav.split('/')[-1].split('.')[0] + '_my',
name='signal_spec-lin_' +
wav.split('/')[-1].split('.')[0] + '_my',
show_plot=show_plot)
plot_spec_profile(x,
feature_extractor.calc_spectogram(x).T,
feature_params['fs'],
feature_extractor.N,
feature_extractor.hop,
log_scale=True,
anno_file=anno,
plot_path=plot_path,
title=wav.split('/')[-1].split('.')[0] + '_my',
name='signal_spec-log_' +
wav.split('/')[-1].split('.')[0] + '_my',
show_plot=show_plot)
# mfcc
if mfcc_plot:
mfcc, bon_pos = feature_extractor.extract_mfcc(
x, reduce_to_best_onset=False)
plot_mfcc_profile(x,
cfg['feature_params']['fs'],
feature_extractor.N,
feature_extractor.hop,
mfcc,
anno_file=anno,
sep_features=True,
bon_pos=bon_pos,
frame_size=cfg['feature_params']['frame_size'],
plot_path=plot_path,
name='signal_mfcc_' +
wav.split('/')[-1].split('.')[0] + '_my',
close_plot=False,
show_plot=show_plot)
def extract_mfcc_data(self, wavs, annos, n_examples, set_name=None):
"""
extract mfcc data from wav-files
wavs must be in a 2D-array [[wavs_class1], [wavs_class2]] so that n_examples will work properly
"""
# mfcc_data: [n x m x l], labels and index
mfcc_data, label_data, index_data = np.empty(shape=(0, self.channel_size, self.feature_size, self.frame_size), dtype=np.float64), [], []
# extract class wavs
for class_wavs, class_annos in zip(wavs, annos):
# class annotation file names extraction
class_annos_file_names = [l + i for f, i, l in [self.file_naming_extraction(a, file_ext='.TextGrid') for a in class_annos]]
# number of class examples
num_class_examples = 0
# run through each example in class wavs
for wav in class_wavs:
# extract file namings
file_name, file_index, label = self.file_naming_extraction(wav, file_ext=self.dataset_cfg['file_ext'])
# get annotation if available
anno = None
if label + file_index in class_annos_file_names: anno = class_annos[class_annos_file_names.index(label + file_index)]
# load and pre-process audio
x, wav_is_useless = self.wav_pre_processing(wav)
if wav_is_useless: continue
# print some info
if self.verbose: print("wav: [{}] with label: [{}], samples=[{}], time=[{}]s".format(wav, label, len(x), len(x) / self.feature_params['fs']))
# extract feature vectors [m x l]
mfcc, bon_pos = self.feature_extractor.extract_mfcc(x, reduce_to_best_onset=False)
# collect wavs
if self.collect_wavs: self.pre_wavs.append((librosa.util.normalize(x), label + str(file_index) + '_' + set_name, bon_pos))
# plot mfcc features
plot_mfcc_profile(x, self.feature_params['fs'], self.feature_extractor.N, self.feature_extractor.hop, mfcc, anno_file=anno, onsets=None, bon_pos=bon_pos, mient=None, minreg=None, frame_size=self.frame_size, plot_path=self.plot_paths['mfcc'], name=label + str(file_index) + '_' + set_name, enable_plot=self.dataset_cfg['enable_plot'])
# damaged file check
if self.dataset_cfg['filter_damaged_files']:
# handle damaged files
if self.detect_damaged_file(mfcc, wav): continue
# add to mfcc_data container
mfcc_data = np.vstack((mfcc_data, mfcc[np.newaxis, :, :, bon_pos:bon_pos+self.frame_size]))
label_data.append(label)
index_data.append(label + file_index)
# update number of examples per class
num_class_examples += 1
# stop if desired examples are reached
if num_class_examples >= n_examples: break
return mfcc_data, label_data, None, index_data
# load audio
x, _ = librosa.load(wav, sr=16000)
# feature extraction
mfcc, bon_pos = feature_extractor.extract_mfcc(x, reduce_to_best_onset=False)
print("mfcc: ", mfcc.shape)
# invert mfcc
x_hat = feature_extractor.invert_mfcc(np.squeeze(mfcc))
# save invert mfcc
soundfile.write(wav.split('.wav')[0] + '_inv_mfcc.wav', x_hat, 16000, subtype=None, endian=None, format=None, closefd=True)
print("x_hat: ", x_hat.shape)
plot_mfcc_profile(x, 16000, feature_extractor.N, feature_extractor.hop, mfcc, anno_file=anno, sep_features=True, diff_plot=False, bon_pos=bon_pos, frame_size=cfg['feature_params']['frame_size'], plot_path=wav_dir, name=wav.split('/')[-1].split('.')[0], show_plot=False, close_plot=False)
plot_waveform(x, 16000, anno_file=anno, hop=feature_extractor.hop, title=wav.split('/')[-1].split('.')[0]+'_my', plot_path=wav_dir, name=wav.split('/')[-1].split('.')[0], show_plot=False)
# random
#x = np.random.randn(16000)
#mfcc, bon_pos = feature_extractor.extract_mfcc(x, reduce_to_best_onset=False)
#plot_mfcc_profile(x, 16000, feature_extractor.N, feature_extractor.hop, mfcc, bon_pos=bon_pos, name='rand', show_plot=True)