def test_pgc1():
assert nb_channel==1
in_buffer = hls.moving_sinus(length, sample_rate=sample_rate, speed = .5, f1=500., f2=2000., ampl = .8)
in_buffer = np.tile(in_buffer[:, None],(1, nb_channel))
loss_params = { 'left' : {'freqs' : [ 125*2**i for i in range(7) ], 'compression_degree': [0]*7, 'passive_loss_db' : [0]*7 } }
loss_params['right'] = loss_params['left']
processing_conf = dict(nb_freq_band=5, level_step=10, debug_mode=True, chunksize=chunksize, backward_chunksize=backward_chunksize, loss_params=loss_params)
processing, online_arrs = hls.run_class_offline(hls.InvCGC, in_buffer, chunksize, sample_rate, processing_conf=processing_conf, buffersize_margin=backward_chunksize)
n = processing.nb_freq_band
in_buffer2 = np.tile(in_buffer,(1, processing.nb_freq_band))
online_arr = online_arrs['pgc1']
offline_arr = in_buffer2.copy()
for i in range(n):
offline_arr[:, i] = scipy.signal.sosfilt(processing.coefficients_pgc[i,:,:], in_buffer2[:,i])
offline_arr = offline_arr[:online_arr.shape[0]]
residual = np.abs((online_arr.astype('float64')-offline_arr.astype('float64'))/np.mean(np.abs(offline_arr.astype('float64'))))
#~ print(np.max(residual))
freq_band = 4
fig, ax = plt.subplots(nrows = 2, sharex=True)
#~ ax[0].plot(in_buffer2[:, freq_band], color = 'k')
ax[0].plot(offline_arr[:, freq_band], color = 'g')
ax[0].plot(online_arr[:, freq_band], color = 'r', ls='--')
ax[1].plot(residual[:, freq_band], color = 'm')
for i in range(nloop):
ax[1].axvline(i*chunksize)
plt.show()
assert np.max(residual)<1e-5, 'pgc1 online differt from offline'
def test_invcomp():
#~ in_buffer = hls.moving_erb_noise(length)
in_buffer = hls.moving_sinus(length, sample_rate=sample_rate, speed = .5, f1=100., f2=2000., ampl = .8)
in_buffer = np.tile(in_buffer[:, None],(1, nb_channel))
#~ print(in_buffer.shape)
#~ exit()
loss_params = { 'left' : {'freqs' : [ 125*2**i for i in range(7) ], 'compression_degree': [0]*7, 'passive_loss_db' : [0]*7 } }
loss_params['right'] = loss_params['left']
processing_conf = dict(nb_freq_band=32, level_step=1., level_max = 100., loss_params=loss_params,
low_freq=100., high_freq=15000.,
debug_mode=True, chunksize=chunksize, backward_chunksize=backward_chunksize)
processing, online_arrs = hls.run_class_offline(hls.InvComp, in_buffer, chunksize, sample_rate, processing_conf=processing_conf, buffersize_margin=backward_chunksize)
print('nlevel', processing.levels.size, 'nb_freq_band', processing.nb_freq_band)
freq_band = 15
fig, ax = plt.subplots(nrows = 7, sharex=True) #, sharey=True)
ax[0].plot(in_buffer[:, 0], color = 'k')
steps = ['pgc1', 'levels', 'dyngain', 'pgc2', 'passive']
for i, k in enumerate(steps):
online_arr = online_arrs[k]
print(online_arr.shape)
ax[i+1].plot(online_arr[:, freq_band], color = 'b')
ax[i+1].set_ylabel(k)
#~ ax[0].plot(offline_arr[:, 0], color = 'g')
out_buffer = online_arrs['main_output']
ax[-1].plot(out_buffer[:, 0], color = 'k')
if nb_channel==2:
#test stereo is like mono
#~ fig, ax = plt.subplots()
#~ ax.plot(out_buffer[:,0], color='b')
#~ ax.plot(out_buffer[:,1], color='r')
#~ fig, ax = plt.subplots()
#~ ax.plot(out_buffer[:,0]-out_buffer[:,1], color='b')
#~ plt.show()
assert np.all(np.abs(out_buffer[:,0]-out_buffer[:,1])<1e-5)
plt.show()
def test_passive_loss():
in_buffer = hls.moving_sinus(length, sample_rate=sample_rate, speed = .5, f1=500., f2=2000., ampl = .8)
in_buffer = np.tile(in_buffer[:, None],(1, nb_channel))
loss_params = { 'left' : {'freqs' : [ 125*2**i for i in range(7) ], 'compression_degree': [1.]*7, 'passive_loss_db' : [-20.]*7 } }
loss_params['right'] = loss_params['left']
processing_conf = dict(nb_freq_band=32, level_step=10, debug_mode=True, chunksize=chunksize, backward_chunksize=backward_chunksize, loss_params=loss_params)
processing, online_arrs = hls.run_class_offline(hls.InvCGC, in_buffer, chunksize, sample_rate, processing_conf=processing_conf, buffersize_margin=backward_chunksize)
n = processing.nb_freq_band
online_pgc2 = online_arrs['pgc2']
online_passive = online_arrs['passive']
#~ offline_passive = online_pgc2.copy()
#~ channels = ('left', 'right')[:nb_channel]
#~ for c, chan in enumerate(channels):
#~ for i in range(n):
#~ offline_passive[:, c*n + i] = processing.passive_gain[c*n + i] * online_pgc2
offline_passive = processing.passive_gain.T * online_pgc2
residual = np.abs((online_passive.astype('float64')-offline_passive.astype('float64'))/np.mean(np.abs(offline_passive.astype('float64'))))
print(np.max(residual))
freq_band = 15
fig, ax = plt.subplots(nrows = 2, sharex=True)
#~ ax[0].plot(in_buffer2[:, freq_band], color = 'k')
ax[0].plot(offline_passive[:, freq_band], color = 'g')
ax[0].plot(online_passive[:, freq_band], color = 'r', ls='--')
ax[1].plot(residual[:, freq_band], color = 'm')
for i in range(nloop):
ax[1].axvline(i*chunksize)
plt.show()
assert np.max(residual)<1e-4, 'passive online differt from offline'
