def relParToWave(x_rel):
global infant
x_abs = get_abs_coord(x_rel)
if infant:
wavFile = parToWave(x_abs, speaker='infant', simulation_name='infant', pitch_var=0.0, len_var=1.0, verbose=False, rank=1, different_folder='aux_infant.wav', monotone=False)
else:
wavFile = parToWave(x_abs, speaker='adult', simulation_name='adult', pitch_var=0.0, len_var=1.0, verbose=False, rank=1, different_folder='aux_adult.wav', monotone=False)
sound = Sound(wavFile)
sound = correct_initial(sound) # call correct_initial to remove initial burst
sound_resampled = get_resampled(sound)
# call get_resampled to adapt generated sound to AN model
sound_extended = get_extended(sound_resampled)
# call get_extended to equalize duration of all sounds
sound_extended.save(wavFile)
return sound_extended
def evaluate_environment(params, i_global, simulation_name, outputfolder, i_target=0, rank=1, speaker='adult', n_vow=5, normalize=False):
folder = outputfolder
############### Sound generation
if output:
print 'simulating vocal tract'
wavFile = parToWave(params, speaker, simulation_name, verbose=output, rank=rank) # call parToWave to generate sound file
# wavFile = par_to_wav(params, speaker, simulation_name, verbose=output, rank=rank) # call parToWave to generate sound file
if output:
print 'wav file '+str(wavFile)+' produced'
sound = loadsound(wavFile) # load sound file for brian.hears processing
if output:
print 'sound loaded'
############### Audio processing
sound = correct_initial(sound) # call correct_initial to remove initial burst
sound_resampled = get_resampled(sound)
# call get_resampled to adapt generated sound to AN model
sound_extended = get_extended(sound_resampled)
# call get_extended to equalize duration of all sounds
sound_extended.save(wavFile) # save current sound as sound file
os.system('cp '+wavFile+' '+folder+'data/vowel_'+str(i_target)+'_'+str(rank)+'.wav')
if playback:
print 'playing back...'
sound_extended.play(sleep=True) # play back sound file
if output:
print 'sound acquired, preparing auditory processing'
out = drnl(sound_extended) # call drnl to get cochlear activation
############### Classifier evaluation
flow_name = 'data/current_auditory_system.flow'
flow_file = open(flow_name, 'r') # open classifier file
flow = cPickle.load(flow_file) # load classifier
flow_file.close() # close classifier file
sample_vote_unnormalized = flow(out) # evaluate trained output units' responses for current item
if normalize:
sample_vote = normalize_activity(sample_vote_unnormalized)
else:
sample_vote = sample_vote_unnormalized
mean_sample_vote = np.mean(sample_vote, axis=0)
# average each output neurons' response over time
confidences = get_confidences(mean_sample_vote)
plot_reservoir_states(flow, sample_vote, i_target, folder, n_vow, rank)
return confidences
def evaluate_environment(params,
i_global,
simulation_name,
outputfolder,
i_target=0,
rank=1,
speaker='adult',
n_vow=5,
normalize=False):
folder = outputfolder
############### Sound generation
if output:
print 'simulating vocal tract'
wavFile = parToWave(params,
speaker,
simulation_name,
verbose=output,
rank=rank) # call parToWave to generate sound file
# wavFile = par_to_wav(params, speaker, simulation_name, verbose=output, rank=rank) # call parToWave to generate sound file
if output:
print 'wav file ' + str(wavFile) + ' produced'
sound = loadsound(wavFile) # load sound file for brian.hears processing
if output:
print 'sound loaded'
############### Audio processing
sound = correct_initial(
sound) # call correct_initial to remove initial burst
sound_resampled = get_resampled(sound)
# call get_resampled to adapt generated sound to AN model
sound_extended = get_extended(sound_resampled)
# call get_extended to equalize duration of all sounds
sound_extended.save(wavFile) # save current sound as sound file
os.system('cp ' + wavFile + ' ' + folder + 'data/vowel_' + str(i_target) +
'_' + str(rank) + '.wav')
if playback:
print 'playing back...'
sound_extended.play(sleep=True) # play back sound file
if output:
print 'sound acquired, preparing auditory processing'
out = drnl(sound_extended) # call drnl to get cochlear activation
############### Classifier evaluation
flow_name = 'data/current_auditory_system.flow'
flow_file = open(flow_name, 'r') # open classifier file
flow = cPickle.load(flow_file) # load classifier
flow_file.close() # close classifier file
sample_vote_unnormalized = flow(
out) # evaluate trained output units' responses for current item
if normalize:
sample_vote = normalize_activity(sample_vote_unnormalized)
else:
sample_vote = sample_vote_unnormalized
mean_sample_vote = np.mean(sample_vote, axis=0)
# average each output neurons' response over time
confidences = get_confidences(mean_sample_vote)
plot_reservoir_states(flow, sample_vote, i_target, folder, n_vow, rank)
return confidences
