dumb spectogram

spect2.py

from scikits.audiolab import Sndfile
import numpy as np
import pylab as plt
import csv
import matplotlib.mlab as mlab
from math import floor, log
#NFFT windows make calculus faster if a power of 2 
closest_2_power=lambda x : 1<<int(floor(log(x,2)))

def load_freq(fn="freq.csv"):
    _csv=csv.reader(open(fn), delimiter='\t')
    note={}
    for i,row in enumerate(_csv):
        if i:
            note[row[0]]=float(row[1])
    return note

koln=Sndfile("test.wav")

note=load_freq()
tempo= 120#tempo in beat per minutes

def ticks_in_range(_min=0,_max=16000,note=note):
    note_name=[]
    note_freq=[]
    for k,v in sorted(note.iteritems()):
        if _max >= v >= _min:
            note_name+=[k]
            note_freq+=[v]
    return note_name, note_freq
    
_min_sec=5
_max_sec=10
_begin_at=_min_sec * koln.samplerate
_stop_at=_max_sec * koln.samplerate

#skipping n seconds
koln.read_frames(_begin_at)

sample=koln.read_frames(_stop_at-_begin_at)
#plt.interactive(True)

## something like the time window used to higher the resolution
## we take 1 measure 4 beats + shannon => * 2 
NFFT=int(8.0* koln.samplerate/float(tempo)) #we want to see the round
## well I let 1 black note overlap
noverlap=NFFT>>2
## A4 flat frequency but I am french we say la not A
la=442.0

## freq lowest limit
lower=13
## A4/2 => A3
higher=la/2

plt.interactive(False)
## testing all colormap because ... it is hard to find the good one
maps=[m for m in plt.cm.datad if not m.endswith("_r")]

for _map in maps:
    print _map
    plt.clf()
    fig=plt.figure( figsize=(30.0,14.0))
    ax=fig.add_subplot(111)
    pxx,  freq, t,cax=ax.specgram(
        ## right chan
        sample[:,0], 
        ## for getting the time scale
        Fs=koln.samplerate, 
        ### Window size
        NFFT=NFFT,
        ## overlap size
        noverlap=noverlap,  
        ### colormap
        cmap=plt.get_cmap(_map))
    ax.set_title(str(_map))
    note_name,note_freq=ticks_in_range(lower,higher)
    ax.set_ylim(lower,higher)
    ax.set_yticks(np.array(note_freq))
    ax.set_yticklabels(note_name)
    step= 1.0 * ( _max_sec - _min_sec ) * tempo / 60
    step=1/step
    ### trying to graph the ticks according to the guessed tempo
    ax.set_xticks(np.arange(0 ,1.0*( _max_sec - _min_sec), step, float))

    ## colorbar is nice
    fig.colorbar(cax)
    ##saving result
    plt.savefig("sono.%s.png" % _map)
    

Ça gère ;)

You have left one line from your code listing:
maps=[m for m in plt.cm.datad if not m.endswith("_r")]

Nice post though, thank you. (I mean this one http://beauty-of-imagination.blogspot.fr/2012/06/color-of-music-why-not-discover-pythons.html)

corrected thx