Re: [Matplotlib-users] Re peated calls to Specgram for animation, memory leaking

[Michael D. <md...@st...>]

Images added to an axes get added to the axes' "images" member, so you 
can simply remove it from there.

I've attached a modified version of your script that does this.  It uses 
a global variable which is probably not best practice, but it should be 
enough to give you the idea.

Cheers,
Mike

xianthax wrote:
> Hello,
>
> First thanks for the great library, recently came across it and seems quite
> useful for what i'm doing.
>
> What i'm trying to do is create an animated specgram().  I'm feeding in
> audio data from a microphone at the moment, although the ultimate use is to
> chart data that will be input via the line in audio port from an FMCW radar
> front end for basic visualization.
>
> I've attached simplified code below based from one of your animation
> examples that shows my problem.  Initially i get fine performance given the
> timeout i've set.  After about a second or two, performance goes down hill
> and progressively gets worse and memory use continues to grow as if each
> call to specgram creates a new instance in memory or worse is drawing a new
> instance over the old.  Is there another way to call specgram to avoid this? 
> It seems the other plot types generally give the ability to update the data
> for an already created graph, i haven't seen a way to do this with specgram,
> perhaps this is what i'm missing?  Ultimately i would like to get the graph
> image out of specgram and append them properly to get a smoother scrolling
> effect but i need to get the data out of specgram fast enough first.
>
> thanks in advance for any pointers.
>
> import gobject
> import numpy as np
> import matplotlib
> import array
> matplotlib.use('GTKAgg')
>
> import matplotlib.pyplot as plt
> import ossaudiodev as oss
>
> audio = oss.open('/dev/dsp','r')
>
> print audio.setfmt(oss.AFMT_S16_LE)
> print audio.channels(1)
> print audio.speed(44100)
>
> fig = plt.figure()
> ax = fig.add_subplot(111)
> data = array.array('H',audio.read(5880))
> img = ax.specgram(data, NFFT=1024,Fs=44100, Fc=0,noverlap=64)
>
> def update():
>     data = array.array('H',audio.read(5880))
>     img = plt.specgram(data, NFFT=1024,Fs=44100, Fc=0,noverlap=64)                      
>     fig.canvas.draw_idle()
>     return True  
>
> gobject.timeout_add(100, update)  
> plt.show()
>
>
>   

-- 
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA