def play_and_analyze(parameters, instrument_name, gesture, plot):
duration = len(parameters)/GESTURE_SAMPLING_FREQUENCY
my_synth = Synth(duration, instrument_name, None, GESTURE_SAMPLING_FREQUENCY)
output_analysis = []
for step_parameters in parameters:
my_synth.set_synthesis_parms(step_parameters)
errcode = my_synth.step_synth()
output_analysis.append(my_synth.get_analysis_values())
output_analysis = np.stack(output_analysis)
my_synth.cleanup()
if plot:
x = np.arange(len(parameters))
gesture_plot_height = int(np.ceil(gesture.shape[1]/2))
total_plot_height = int(gesture_plot_height + 4)
# Find the most similar
simils = []
for g_axis in gesture.T:
simils.append([ mse(g_axis, feature) for feature in output_analysis.T ])
most_simil = [ np.argmin(s) for s in simils ]
# Plot gesture
iv, jv = np.meshgrid(np.arange(gesture_plot_height), np.arange(2), indexing='ij')
plot_coords = list(zip(np.ndarray.flatten(iv), np.ndarray.flatten(jv)))
plot_shape = (total_plot_height, 2)
for k, g_axis in enumerate(gesture.T):
ax = plt.subplot2grid(plot_shape, plot_coords[k])
ax.plot(x, g_axis, label='gesture axis {}'.format(k), color=colors[k])
ax.legend(loc='upper right')
ax.set_ylim(0,1)
# Plot audio features
iv, jv = np.meshgrid(np.arange(gesture_plot_height, total_plot_height),
np.arange(2), indexing='ij')
plot_coords = list(zip(np.ndarray.flatten(iv), np.ndarray.flatten(jv)))
for k, feature in enumerate(output_analysis.T):
color = 'b'
if k in most_simil:
if most_simil.count(k) > 1:
color = 'k'
else:
color = colors[most_simil.index(k)]
audio_features = ['amp', 'env_crest', 'pitch', 'centroid',
'flatness', 's_crest', 'flux', 'mfcc_diff']
ax = plt.subplot2grid(plot_shape, plot_coords[k])
ax.plot(x, feature, color=color, label=audio_features[k])
ax.legend(loc='upper right')
ax.set_ylim(0,1)
similarity = np.sum([ min(s) for s in simils ])
plt.savefig('/shape/sounds/{}.png'.format(my_synth.filename), dpi=300)
plt.close()
return (my_synth.filename, similarity)
