improve audio timeline, normalize to full width

This commit is contained in:
j 2012-03-07 15:36:48 +01:00
parent efc81d7bc9
commit 0f3045f829

View file

@ -13,8 +13,7 @@ import gst
class Audio(gst.Pipeline): class Audio(gst.Pipeline):
left = [] volume = []
right = []
position = 0 position = 0
def __init__(self, uri, channels=2): def __init__(self, uri, channels=2):
@ -34,12 +33,12 @@ class Audio(gst.Pipeline):
self.queue = gst.element_factory_make("queue") self.queue = gst.element_factory_make("queue")
self.rate = gst.element_factory_make("audioresample") self.rate = gst.element_factory_make("audioresample")
self.conv = gst.element_factory_make("audioconvert") self.conv = gst.element_factory_make("audioconvert")
self.level = gst.element_factory_make("level")
self.level.props.interval = int(gst.SECOND / self.framerate)
self.sink = gst.element_factory_make("fakesink") self.sink = gst.element_factory_make("fakesink")
self.sink.props.signal_handoffs = True
self.sink.connect('handoff', self._data_callback)
self.add(self.src, self.sbin, self.conv, self.queue, self.rate, self.sink) self.add(self.src, self.sbin, self.conv, self.queue, self.rate, self.level, self.sink)
self.src.link(self.sbin) self.src.link(self.sbin)
@ -61,61 +60,10 @@ class Audio(gst.Pipeline):
self.conv.link(self.rate, self.conv.link(self.rate,
gst.Caps("audio/x-raw-int,channels=%s,width=16,depth=16" % gst.Caps("audio/x-raw-int,channels=%s,width=16,depth=16" %
(self.channels, ))) (self.channels, )))
self.rate.link(self.sink, self.rate.link(self.level,
gst.Caps("audio/x-raw-int,rate=%s,channels=%s,width=16,depth=16" % gst.Caps("audio/x-raw-int,rate=%s,channels=%s,width=16,depth=16" %
(self.samplerate, self.channels))) (self.samplerate, self.channels)))
self.level.link(self.sink)
def _data_callback(self, sink, buff, pad):
timestamp = buff.timestamp
samples = int(buff.size // 2 / self.channels)
fmt = "<" + str(samples) + "h"
data = unpack(fmt, buff.data[:2*samples]) + unpack(fmt, buff.data[2*samples:])
n = 0
for i in data:
if n%2 == 0:
self.left.append(i)
else:
self.right.append(i)
n += 1
samples_per_pixel = int(self.samplerate / self.framerate)
while len(self.left) > samples_per_pixel:
tile = int(math.floor(float(self.position) / self.input_tile_width))
tilePos = int(self.position - (tile * self.input_tile_width))
def plot(p, start):
p = int(p / 256)
height = int((p * self.tile_height) / 256) * 2
crop = int((self.tile_height-height) / 2)
if p: p += 20
color = (p, p, p)
color = (255,255,255)
if start:
end = self.tile_height - crop
else:
start = crop
end = int(self.tile_height/2)
#for i in range(0, self.tile_height):
for i in range(start, end):
self.tiles[tile].putpixel((tilePos, i), color)
#left
pixel = np.asarray(self.left[:samples_per_pixel])
self.left = self.left[samples_per_pixel:]
p = np.sum(np.abs(pixel)) / samples_per_pixel
plot(p, 0)
#right
pixel = np.asarray(self.right[:samples_per_pixel])
self.right = self.right[samples_per_pixel:]
p = np.sum(np.abs(pixel)) / samples_per_pixel
plot(p, int(self.tile_height/2))
self.position += 1
if self.mainloop and timestamp >= self.duration:
self.done()
def getDuration(self): def getDuration(self):
if self.duration < 0: if self.duration < 0:
@ -126,6 +74,21 @@ class Audio(gst.Pipeline):
format, self.duration = q.parse_duration() format, self.duration = q.parse_duration()
return self.duration return self.duration
def done(self):
pass
def onBusMessage(self, bus, message):
if message.src == self.level:
struct = message.structure
if struct.get_name() == 'level':
self.volume.append((
pow (10, struct['rms'][0] / 20),
pow (10, struct['rms'][1] / 20)
))
if message.src == self and message.type == gst.MESSAGE_EOS:
self.done()
class Timeline(Audio): class Timeline(Audio):
def __init__(self, uri, prefix, width, height): def __init__(self, uri, prefix, width, height):
Audio.__init__(self, uri) Audio.__init__(self, uri)
@ -150,6 +113,31 @@ class Timeline(Audio):
self.set_state(gst.STATE_PLAYING) self.set_state(gst.STATE_PLAYING)
self.mainloop.run() self.mainloop.run()
normalize = 1/max([max(*v) for v in self.volume])
for volume in self.volume:
tile = int(math.floor(float(self.position) / self.input_tile_width))
tilePos = int(self.position - (tile * self.input_tile_width))
def plot(p, start):
height = normalize * p * self.tile_height
crop = int((self.tile_height-height) / 2)
color = (p, p, p)
color = (255,255,255)
if start:
end = self.tile_height - crop
else:
start = crop
end = int(self.tile_height/2)
for i in range(start, end):
self.tiles[tile].putpixel((tilePos, i), color)
#left
plot(volume[0], 0)
#right
plot(volume[1], int(self.tile_height/2))
self.position += 1
for i in range(ntiles): for i in range(ntiles):
tile = self.tiles[i] tile = self.tiles[i]
if tile.size[0] != self.tile_width: if tile.size[0] != self.tile_width:
@ -167,8 +155,3 @@ class Timeline(Audio):
def done(self): def done(self):
self.mainloop.quit() self.mainloop.quit()
def onBusMessage(self, bus, message):
if message.src == self and message.type == gst.MESSAGE_EOS:
self.done()