Slengths)
num_plots = len(spec_templates)
num_rows = 2
num_cols = num_plots/num_rows+1
for i in xrange(len(spec_templates)):
plt.subplot(num_cols,num_rows,i+1)
plt.imshow(spec_templates[i].T[::-1],interpolation='nearest')
plt.savefig('aar1_spec_templates_%d.png' % num_mix)
np.savez('aar1_templates_%d.npz' % num_mix, templates)
np.save('aar1_affinities_%d.npy' % num_mix, bem.affinities)
test_path = '/home/mark/Template-Speech-Recognition/Data/Test/'
test_file_indices = gtrd.get_data_files_indices(test_path)
test_example_lengths = gtrd.get_detect_lengths(test_file_indices,test_path)
np.save("data/test_example_lengths.npy",test_example_lengths)
np.save("data/test_file_indices.npy",test_file_indices)
import collections
FOMS = collections.defaultdict(list)
bgd = np.clip(np.load("data/aar_bgd_mel.npy"),.1,.4)
test_example_lengths =np.load("data/test_example_lengths.npy")
test_path = "/home/mark/Template-Speech-Recognition/Data/Test/"
test_file_indices =np.load("data/test_file_indices.npy")
for num_mix in num_mix_params:
templates = tuple(np.clip(T,.01,.99) for T in (np.load('aar1_templates_%d.npz' % num_mix))['arr_0'])
detection_array = np.zeros((test_example_lengths.shape[0],
test_example_lengths.max() + 2),dtype=np.float32)
def get_params(
args,
sample_rate=16000,
num_window_samples=320,
num_window_step_samples=80,
fft_length=512,
kernel_length=7,
freq_cutoff=3000,
use_mel=False,
):
# we get the basic file paths right here
# TODO: make this system adaptive
root_path = "/home/mark/Template-Speech-Recognition/"
utterances_path = "/home/mark/Template-Speech-Recognition/Data/Train/"
try:
file_indices = np.load("data_parts/train_file_indices.npy")
except:
file_indices = gtrd.get_data_files_indices(utterances_path)
np.save("data_parts/train_file_indices.npy", file_indices)
num_mix_params = [1, 2, 3, 5, 7, 9]
test_path = "/home/mark/Template-Speech-Recognition/Data/Test/"
train_path = "/home/mark/Template-Speech-Recognition/Data/Train/"
try:
test_example_lengths = np.load("data_parts/test_example_lengths.npy")
test_file_indices = np.load("data_parts/test_file_indices.npy")
except:
test_file_indices = gtrd.get_data_files_indices(test_path)
test_example_lengths = gtrd.get_detect_lengths(test_file_indices, test_path)
np.save("data_parts/test_example_lengths.npy", test_example_lengths)
np.save("data_parts/test_file_indices.npy", test_file_indices)
try:
train_example_lengths = np.load("data_parts/train_example_lengths.npy")
train_file_indices = np.load("data_parts/train_file_indices.npy")
except:
train_file_indices = gtrd.get_data_files_indices(train_path)
train_example_lengths = gtrd.get_detect_lengths(train_file_indices, train_path)
np.save("data_parts/train_example_lengths.npy", train_example_lengths)
np.save("data_parts/train_file_indices.npy", train_file_indices)
return (
gtrd.SpectrogramParameters(
sample_rate=16000,
num_window_samples=320,
num_window_step_samples=80,
fft_length=512,
kernel_length=7,
freq_cutoff=3000,
use_mel=args.use_mel,
),
gtrd.makeEdgemapParameters(
block_length=args.edgeMapBlockLength,
spread_length=args.edgeMapSpreadLength,
threshold=args.edgeMapThreshold,
),
root_path,
utterances_path,
file_indices,
num_mix_params,
test_path,
train_path,
train_example_lengths,
train_file_indices,
test_example_lengths,
test_file_indices,
)
file_indices,
syllable,
num_examples=-1,
verbose=True)
clipped_bgd = np.clip(avg_bgd.E,.1,.4)
np.save(tmp_data_path+'clipped_bgd_102012.npy',clipped_bgd)
padded_examples, lengths = et.extend_examples_to_max(clipped_bgd,syllable_examples,
return_lengths=True)
aar_template, aar_registered = et.register_templates_time_zero(syllable_examples,min_prob=.01)
test_example_lengths = gtrd.get_detect_lengths(data_path+'Test/')
np.save(tmp_data_path+'test_example_lengths_102012.npy',test_example_lengths)
detection_array = np.zeros((test_example_lengths.shape[0],
test_example_lengths.max() + 2),dtype=np.float32)
linear_filter,c = et.construct_linear_filter(aar_template,
clipped_bgd)
# need to state the syllable we are working with
syllable = np.array(['aa','r'])
detection_array,example_start_end_times, detection_lengths = gtrd.get_detection_scores(data_path+'Test/',
detection_array,
Slengths)
num_plots = len(spec_templates)
num_rows = 2
num_cols = num_plots/num_rows+1
for i in xrange(len(spec_templates)):
plt.subplot(num_cols,num_rows,i+1)
plt.imshow(spec_templates[i].T[::-1],interpolation='nearest')
plt.savefig('aar1_spec_templates_%d.png' % num_mix)
np.savez('aar1_templates_%d.npz' % num_mix, templates)
np.save('aar1_affinities_%d.npy' % num_mix, bem.affinities)
train_path = '/home/mark/Template-Speech-Recognition/Data/Train/'
train_file_indices = gtrd.get_data_files_indices(train_path)
train_example_lengths = gtrd.get_detect_lengths(train_file_indices,train_path)
np.save("data/train_example_lengths.npy",train_example_lengths)
np.save("data/train_file_indices.npy",train_file_indices)
import collections
FOMS = collections.defaultdict(list)
for num_mix in num_mix_params:
templates = (np.load('aar1_templates_%d.npz' % num_mix))['arr_0']
detection_array = np.zeros((train_example_lengths.shape[0],
train_example_lengths.max() + 2),dtype=np.float32)
linear_filters_cs = et.construct_linear_filters(templates,
bgd)
np.savez('data/linear_filter_aar_%d.npy'% num_mix,linear_filters_cs[:][0])
np.savez('data/c_aar_%d.npy'%num_mix,np.array(linear_filters_cs[:][1]))
syllable = np.array(['aa','r'])
detection_array,example_start_end_times, detection_lengths = gtrd.get_detection_scores_mixture(train_path,
