def __init__(self, trainer):
Process.__init__(self)
self.trainer = trainer
self.queue = Queue(1)
self.network_output_rms = RMS(50)
class Plot(Process):
def __init__(self, trainer):
Process.__init__(self)
self.trainer = trainer
self.queue = Queue(1)
self.network_output_rms = RMS(50)
def update(self):
try:
data = Data()
data.time = self.trainer.time
data.frames = self.trainer.inputs.frames
data.mfcc = \
self.trainer.network.capture_transformed_inputs(13)
data.output = self.trainer.outputs.value
data.network_output = self.trainer.network.capture_output(1)[0]
self.network_output_rms.append(data.network_output)
data.network_output_rms = self.network_output_rms()
self.queue.put_nowait(data)
except Full: pass
def run(self):
self.figure = pyplot.figure()
self.plot_audio = pyplot.subplot(311)
self.plot_mfcc = pyplot.subplot(312)
self.plot_output = pyplot.subplot(313)
self.audio_line, = self.plot_audio.plot([], [])
self.plot_audio.set_ylim(-0.2, 0.2)
self.plot_audio.grid(True)
self.image_mfcc = self.plot_mfcc.imshow(
numpy.empty([13, SIM_STEPS_PER_WINDOW]),
interpolation="gaussian", aspect="auto")
self.image_mfcc.set_extent((0, 100, 13, 0))
self.image_mfcc.set_clim(-0.1, 0.1)
self.line_output, = self.plot_output.plot([], [])
self.line_network_output, = self.plot_output.plot([], [])
self.line_network_output_rms, = self.plot_output.plot([], [])
self.plot_output.set_ylim(-0.1, 1.1)
self.plot_output.grid(True)
self.time_data = deque()
self.audio_data = deque()
self.output_time_data = deque()
self.output_data = deque()
self.network_output_data = deque()
self.network_output_rms_data = deque()
self.mfcc_data = numpy.zeros([13, SIM_STEPS_PER_WINDOW])
animation_object = animation.FuncAnimation(self.figure,
lambda frame_id: self.animation(), interval=30)
pyplot.show()
def animation(self):
data = self.queue.get_nowait()
time = data.time
audio_frame_time = SIM_STEP / len(data.frames)
plot_audio_frames_count = PLOT_TIME_WINDOW / \
(audio_frame_time * PLOT_EACH_NTH_AUDIO_FRAME)
for audio_frame in data.frames[::PLOT_EACH_NTH_AUDIO_FRAME]:
self.time_data.append(time)
if len(self.time_data) > plot_audio_frames_count:
self.time_data.popleft()
time += audio_frame_time * PLOT_EACH_NTH_AUDIO_FRAME
self.audio_data.append(audio_frame)
if len(self.audio_data) > plot_audio_frames_count:
self.audio_data.popleft()
self.mfcc_data = numpy.roll(self.mfcc_data, -1)
self.mfcc_data[:,-1] = data.mfcc
self.audio_line.set_data(self.time_data, self.audio_data)
self.plot_audio.set_xlim(
time - PLOT_TIME_WINDOW, time)
self.image_mfcc.set_array(self.mfcc_data)
self.image_mfcc.set_extent(
(time - PLOT_TIME_WINDOW, time, 13, 0))
self.output_time_data.append(data.time)
if len(self.output_time_data) > 100:
self.output_time_data.popleft()
self.output_data.append(data.output)
if len(self.output_data) > 100:
self.output_data.popleft()
self.line_output.set_data(
self.output_time_data, self.output_data)
self.plot_output.set_xlim(data.time - PLOT_TIME_WINDOW, data.time)
self.network_output_data.append(data.network_output)
if len(self.network_output_data) > 100:
self.network_output_data.popleft()
self.line_network_output.set_data(
self.output_time_data, self.network_output_data)
self.network_output_rms_data.append(data.network_output_rms)
if len(self.network_output_rms_data) > 100:
self.network_output_rms_data.popleft()
self.line_network_output_rms.set_data(
self.output_time_data, self.network_output_rms_data)
