def plot_result():
cp_all = np.zeros((n_room, n_mic_pos, n_azi_tar, n_n_inter, n_test))
rmse_all = np.zeros((n_room, n_mic_pos, n_azi_tar, n_n_inter, n_test))
for room_i, room in enumerate(room_all):
for mic_pos_i, mic_pos in enumerate(mic_pos_all):
result_fpath = '../result/{}_{}.npy'.format(room, mic_pos)
[cp_all[room_i, mic_pos_i],
rmse_all[room_i, mic_pos_i]] = np.load(result_fpath)
fig, ax = plt.subplots(1,
2,
figsize=[8, 3],
sharex=True,
tight_layout=True)
for n_inter_i in range(n_n_inter):
ax[0].plot(range(n_room),
np.mean(rmse_all[:, :, :, n_inter_i, :], axis=(1, 2, 3)),
alpha=0.5,
label=f'{n_inter_i+1} src')
ax[1].plot(
range(n_room),
100 * (1 - np.mean(cp_all[:, :, :, n_inter_i, :], axis=(1, 2, 3))),
alpha=0.5,
label=f'{n_inter_i+1} src')
ax[0].plot(range(n_room),
np.mean(rmse_all, axis=(1, 2, 3, 4)),
alpha=1,
color='black',
label=f'mean')
ax[0].xaxis.set_ticks(range(n_room))
ax[0].set_xticklabels([room[3:] for room in room_all])
ax[0].set_xlabel('Room')
ax[0].set_ylabel('RMSE($^o$)')
# ax[0].spines['top'].set_visible(False)
ax[1].plot(range(n_room),
100 * (1 - np.mean(cp_all, axis=(1, 2, 3, 4))),
alpha=1,
color='black',
label=f'mean')
ax[1].legend()
ax[1].set_xlabel('Room')
ax[1].set_ylim([0, 60])
ax[1].set_ylabel('Anomalies(%)')
plot_tools.savefig(fig,
fig_name='result_room.png',
fig_dir='../images/result')
with open('../result/result.txt', 'w') as result_file:
result_file.write('RMSE')
result_file.write('\t {}'.format(np.mean(rmse_all, axis=(1, 2, 3, 4))))
result_file.write('Anomalies')
result_file.write('\t {}'.format(1 -
np.mean(cp_all, axis=(1, 2, 3, 4))))
def k_means_test():
x, Y = make_blobs(cluster_std=1,
random_state=np.random.randint(100),
n_samples=500,
centers=3)
# Stratch dataset to get ellipsoid data
# x = np.dot(x,np.random.RandomState(0).randn(2,2))
centers, fig = k_means(x, k=3, is_plot=True)
plot_tools.savefig(fig)
def plot_train_process(model_dir):
fig, ax = plt.subplots(1, 1)
plot_settings = {'linewidth': 4}
n_epoch_max = 0
for room_i, room in enumerate(reverb_room_all):
record_fpath = os.path.join(model_dir, room, 'train_record.npz')
record_info = np.load(record_fpath)
cost_record_valid = record_info['cost_record_valid']
# rmse_record_valid = record_info['azi_rmse_record_valid']
n_epoch = np.nonzero(cost_record_valid)[0][-1] + 1
if n_epoch > n_epoch_max:
n_epoch_max = n_epoch
ax.plot(cost_record_valid[:n_epoch], **plot_settings, label=room[-1])
ax.xaxis.set_major_locator(MaxNLocator(integer=True))
ax.legend()
ax.set_ylabel('Cross entrophy')
ax.set_xlabel('Epoch(n)')
plot_tools.savefig(fig, 'learning_curve_all.png', '../images')
std_37d]] = cal_norm_coef(dataset_room_dir_all)
param_name = 'mct_{}'.format(room_tar)
# np.save(os.path.join(norm_param_dir, '{}_37d.npy'.format(param_name)),
# [mean_37d, std_37d])
# np.save(os.path.join(norm_param_dir, '{}_1d.npy'.format(param_name)),
# [mean_1d, std_1d])
ax.errorbar(x=tau + (room_i - 1) * bar_width,
y=mean_37d,
yerr=std_37d,
label=room_tar)
ax.set_xlabel('Delay(sample)')
ax.legend()
plot_tools.savefig(fig, fig_name='norm_coef.png', fig_dir='../images')
# normalization parameters of all train data
if True:
dataset_room_dir_all = [
os.path.join(dataset_dir, room) for room in room_all
]
[[mean_1d, std_1d], [mean_37d,
std_37d]] = cal_norm_coef(dataset_room_dir_all)
np.save(os.path.join(norm_param_dir, 'all_room_37d.npy'),
[mean_37d, std_37d])
np.save(os.path.join(norm_param_dir, 'all_room_1d.npy'),
[mean_1d, std_1d])
is_middle_ear=True,
ihc_type='Roman')
t_frame = np.arange(n_frame) * shift_len + int(frame_len / 2)
fig = plt.figure(figsize=(8, 4), tight_layout=True)
ax1 = plt.subplot(221)
ax1.plot(t_frame, flag_frame_all, color='red', label='Criteria')
ax1.legend(loc='upper right')
ax_twin = ax1.twinx()
ax_twin.plot(np.sum(front_end.filter(tar_record)[band_i], axis=1),
label='tar')
ax_twin.plot(np.sum(front_end.filter(inter_record)[band_i], axis=1),
label='inter')
ax_twin.legend(loc='lower right')
ax2 = plt.subplot(223)
ax2.plot(t_frame, vad_flag_frame_all + 0.09, label='vad')
ax2.plot(t_frame, snr_flag_frame_all + 0.06, label='snr')
ax2.plot(t_frame, ccf_flag_frame_all + 0.03, label='ccf')
ax2.plot(t_frame, itd_flag_frame_all, label='itd')
ax2.legend()
ax3 = plt.subplot(122)
plot_cue_sample(cue_frame_all[band_i], ax3, label='all')
plot_cue_sample(cue_frame_all[band_i, flag_frame_all, :],
ax3,
label='preserve')
ax3.legend()
plot_tools.savefig(fig,
fig_name=f'criterias_eg_{band_i}.png',
fig_dir='../images/train')
def file_reader(fea_dir,
band_tar=None,
azi_tar=None,
is_screen=False,
record_dir=None,
is_plot=False,
fig_name=None,
is_pb=False):
#
theta_vad = 40
theta_corr_coef = 0.3
theta_itd = 44.0 / 44.1
if is_screen:
src_fpath_all = load_src_fpath(record_dir)
fea_fpath_all = get_fpath(fea_dir, suffix='.npz', is_absolute=True)
pb = ProcessBar(len(fea_fpath_all))
for fea_fpath in fea_fpath_all:
if is_pb:
pb.update()
*_, room, mic_pos, fname = fea_fpath[:-4].split('/')
azi, wav_i, snr = [np.int16(item) for item in fname.split('_')]
if (azi_tar is not None) and (azi != azi_tar):
continue
fea_file = np.load(fea_fpath)
cue_frame_all = fea_file['cue_frame_all']
ccf_frame_all = fea_file['ccf_frame_all']
snr_frame_all = fea_file['snr_frame_all']
if not is_screen:
if band_tar is None:
yield np.transpose(cue_frame_all, axes=(1, 0, 2))
else:
yield cue_frame_all[band_tar]
else:
n_frame = cue_frame_all.shape[1]
flag_frame_all_band_all = []
# feature selection
# 1. vad, on one channel(L)
src_fpath_tar = \
src_fpath_all[room][mic_pos][f'{azi}_{wav_i}_{snr}'][0]
src_fpath_tar = src_fpath_tar.replace('Data/TIMIT',
'Data/TIMIT_wav')
src_tar, fs = wav_tools.read_wav(src_fpath_tar)
tar_fpath = ''.join((f'{record_dir}/{room}/{mic_pos}/',
f'{azi}_{wav_i}_{snr}_tar.wav'))
tar, fs = wav_tools.read_wav(tar_fpath)
if tar.shape[0] != src_tar.shape[0]:
raise Exception()
# time delay between source and recording, about 70 samples
delay = 190
src_tar = np.concatenate((src_tar[delay:], np.zeros(delay)))
vad_flag_frame_all = wav_tools.vad(x=src_tar,
frame_len=frame_len,
shift_len=shift_len,
theta=theta_vad,
is_plot=False)
vad_flag_frame_all = np.squeeze(vad_flag_frame_all[:n_frame])
for band_i in range(n_band):
if band_tar is not None:
if band_i != band_tar:
continue
# 2. SNR in each frequency band
snr_flag_frame_all = snr_frame_all[band_i] > 0.0
# 3. correlation coefficients
ccf_flag_frame_all = np.greater(
np.max(ccf_frame_all[band_i], axis=1), theta_corr_coef)
# 4. ITDs range
itd_flag_frame_all = np.less(
np.abs(cue_frame_all[band_i, :, 0]), theta_itd) # itd ms
# combine all criteras
flag_frame_all = np.logical_and.reduce(
(vad_flag_frame_all, snr_flag_frame_all,
ccf_flag_frame_all, itd_flag_frame_all))
flag_frame_all_band_all.append(flag_frame_all)
# plot waveform and corresponding criteria result
if is_plot:
tar_fpath = os.path.join('{}_{}.wav'.format(azi, wav_i))
tar, fs = wav_tools.read_wav(tar_fpath)
inter_fpath = '{}_{}_{}.wav'.format(azi, wav_i, snr)
inter, fs = wav_tools.read_wav(inter_fpath)
front_end = Auditory_model(fs=fs,
cf_low=freq_low,
cf_high=freq_high,
n_band=n_band,
is_middle_ear=True,
ihc_type='Roman')
t_frame = np.arange(n_frame) * shift_len + int(frame_len / 2)
fig = plt.figure(figsize=(8, 4), tight_layout=True)
ax1 = plt.subplot(221)
ax1.plot(np.sum(front_end.filter(inter)[band_i], axis=1))
ax1.plot(np.sum(front_end.filter(tar)[band_i], axis=1))
ax_twin = ax1.twinx()
ax_twin.plot(t_frame, flag_frame_all, color='red')
ax2 = plt.subplot(223)
ax2.plot(t_frame, vad_flag_frame_all + 0.09, label='vad')
ax2.plot(t_frame, snr_flag_frame_all + 0.06, label='snr')
ax2.plot(t_frame, ccf_flag_frame_all + 0.03, label='ccf')
ax2.plot(t_frame, itd_flag_frame_all, label='itd')
ax2.legend()
ax3 = plt.subplot(122)
plot_cue_sample(cue_frame_all[band_i], ax3)
plot_cue_sample(cue_frame_all[band_i, flag_frame_all, :], ax3)
plot_tools.savefig(fig, fig_name=fig_name, fig_dir='./')
return
if band_i is None:
flag_frame_all = np.logical_and.reduce(flag_frame_all_band_all)
yield np.transpose(cue_frame_all[:, flag_frame_all, :],
axes=(1, 0, 2))
else:
flag_frame_all = flag_frame_all_band_all[0]
yield cue_frame_all[band_tar, flag_frame_all, :]
record_fpath = f'{model_dir_base}/{room}/train_record.npz'
record_info = np.load(record_fpath)
cost_record_valid = record_info['cost_record_valid']
n_epoch = np.nonzero(cost_record_valid)[0][-1] + 1
if n_epoch > n_epoch_max:
n_epoch_max = n_epoch
ax[0].plot(cost_record_valid[:n_epoch],
**plot_settings,
label=room[-1])
# rmse
rmse_all = np.load(f'{model_dir_base}/result.npy')
rmse_mean_all = np.mean(rmse_all, axis=0)
rmse_std_all = np.std(rmse_all, axis=0)
ax[0].xaxis.set_major_locator(MaxNLocator(integer=True))
ax[0].legend()
ax[0].set_ylabel('Cross entrophy')
ax[0].set_xlabel('Epoch(n)')
ax[0].set_title('Learning Curve')
ax[1].bar([room[-1] for room in reverb_room_all],
rmse_mean_all,
yerr=rmse_std_all / np.sqrt(n_test))
plt.xticks(rotation=45)
ax[1].set_xlabel('Room')
ax[1].set_ylabel('RMSE')
ax[1].set_title('Test')
man_std = [[rmse_mean_all[i], rmse_std_all[i]] for i in range(4)]
plot_tools.savefig(fig, 'result.png', model_dir_base)