mirror of
https://github.com/brmlab/ledbar.git
synced 2025-06-09 05:14:03 +02:00
192 lines
5.6 KiB
Python
Executable file
192 lines
5.6 KiB
Python
Executable file
#!/usr/bin/python
|
|
# vim:et:sw=4:ts=4:sts=4
|
|
|
|
import pyaudio
|
|
import struct
|
|
import math
|
|
import numpy as np
|
|
import time
|
|
import sys
|
|
import getopt
|
|
from datetime import datetime
|
|
import logging
|
|
import ledbar
|
|
|
|
CHUNK_SIZE = 256
|
|
FORMAT = pyaudio.paInt16
|
|
CHANNELS = 1
|
|
RATE = 44100
|
|
|
|
PIXELS = 20
|
|
LAZY = 0
|
|
SYMMETRIC = 0
|
|
|
|
HISTORY_SIZE = 8
|
|
MIN_FREQ = 50
|
|
MAX_FREQ = 12000
|
|
|
|
ATTENUATION = 10**(40/10) # attenuation of 40dB
|
|
|
|
HUE = 0 # 1 - reddish, 0 - blueish
|
|
|
|
logging.basicConfig(level='WARNING')
|
|
|
|
def print_usage():
|
|
print '''\
|
|
USAGE:
|
|
%s [-l] [-n number] [-s] [-h]
|
|
OPTIONS:
|
|
-l lazy mode
|
|
-n number number of controlled boxes
|
|
-s symmetric mode
|
|
-a number attenuation in dB (try -a40.0)
|
|
-H number hue mode: 0 == blue-green (default), 1 == red-blue
|
|
-h --help show this help
|
|
''' % sys.argv[0]
|
|
|
|
try:
|
|
opts, args = getopt.getopt(sys.argv[1:], 'n:lsha:H:', ['help'])
|
|
except getopt.GetoptError:
|
|
print_usage()
|
|
sys.exit(1)
|
|
if len(args):
|
|
print_usage()
|
|
sys.exit(1)
|
|
for k, v in opts:
|
|
if k == '-n':
|
|
if not v.isdigit():
|
|
print_usage()
|
|
sys.exit(1)
|
|
PIXELS = int(v)
|
|
elif k == '-l':
|
|
LAZY = 1
|
|
elif k == '-s':
|
|
SYMMETRIC = 1
|
|
elif k == '-h' or k == '--help':
|
|
print_usage()
|
|
sys.exit(0)
|
|
elif k == '-a':
|
|
try: v = float(v)
|
|
except ValueError:
|
|
print 'error: attenuation must be float value'
|
|
print_usage()
|
|
ATTENUATION = 10**(v/10)
|
|
elif k == '-H':
|
|
if not v.isdigit():
|
|
print_usage()
|
|
sys.exit(1)
|
|
HUE = int(v)
|
|
|
|
if LAZY == 1:
|
|
HISTORY_SIZE = 12
|
|
if SYMMETRIC == 1:
|
|
EPIXELS = PIXELS / 2
|
|
else:
|
|
EPIXELS = PIXELS
|
|
# EPIXELS: Effective pixels (for spectrum display)
|
|
|
|
SAMPLE_SIZE = CHUNK_SIZE*HISTORY_SIZE
|
|
FREQ_STEP = float(RATE) / (CHUNK_SIZE * HISTORY_SIZE)
|
|
PIXEL_FREQ_RANGE = math.pow(float(MAX_FREQ) / MIN_FREQ, 1.0/EPIXELS)
|
|
|
|
def with_stream( fnc ):
|
|
|
|
p = pyaudio.PyAudio()
|
|
|
|
stream = p.open(format = FORMAT,
|
|
channels = CHANNELS,
|
|
rate = RATE,
|
|
input = True,
|
|
frames_per_buffer = CHUNK_SIZE)
|
|
try:
|
|
fnc(stream)
|
|
finally:
|
|
stream.close()
|
|
p.terminate()
|
|
|
|
|
|
def get_color(volume):
|
|
vol_thres = 200
|
|
if volume <= vol_thres: return (0, 0, 0)
|
|
p = 1-25/(volume-vol_thres)
|
|
if p <= 0: return (0, 0, 0)
|
|
if p >= 1: return (1.0, 1.0, 1.0)
|
|
# Monochromatic mode:
|
|
#p = p * p * p * p * p * p * p
|
|
#return (p, p, 0) # or any other combination
|
|
if LAZY == 1:
|
|
p *= p
|
|
else:
|
|
p *= p * p
|
|
if HUE:
|
|
if p <= 0.4: return (p*2.5,0,0)
|
|
elif p <= 0.7: return (1.0-(p-0.4)*3.33, 0, (p-0.4)*3.33)
|
|
elif p <= 0.9: return (1.0-(p-0.7)*5.0, 0, (p-0.7)*5.0)
|
|
else: return (1.0, (p-0.9)*10.0, (p-0.9)*10.0)
|
|
else:
|
|
if p <= 0.4: return (0, 0, p*2.5)
|
|
elif p <= 0.7: return (0, (p-0.4)*3.33, 1.0-(p-0.4)*3.33)
|
|
elif p <= 0.9: return ((p-0.7)*5.0, 1.0-(p-0.7)*5.0, 0.0)
|
|
else: return (1.0, (p-0.9)*10.0, (p-0.9)*10.0)
|
|
|
|
def loop( stream ):
|
|
l = ledbar.Ledbar(PIXELS)
|
|
history = []
|
|
|
|
history_diminish = np.array([[((i+1.0) / HISTORY_SIZE)**2] * CHUNK_SIZE for i in xrange(HISTORY_SIZE)])
|
|
window = np.array([0.5*(1-math.cos(2*math.pi*i/(SAMPLE_SIZE-1))) for i in xrange(SAMPLE_SIZE)])
|
|
work = True
|
|
|
|
nexttrig = 0
|
|
|
|
while work:
|
|
try: data = stream.read(CHUNK_SIZE)
|
|
except IOError: continue
|
|
nowtrig = datetime.now().microsecond / 50000
|
|
if (nowtrig == nexttrig):
|
|
continue
|
|
else:
|
|
nexttrig = nowtrig
|
|
if len(data) == 0: break
|
|
indata = np.array(struct.unpack('%dh'%CHUNK_SIZE,data))
|
|
history.append(indata)
|
|
if len(history) > HISTORY_SIZE: history.pop(0)
|
|
elif len(history) < HISTORY_SIZE: continue
|
|
|
|
# obtain input sequence ~~ oohhh what a kind of dimmish magic and windowing
|
|
#TODO: dynamic attenuation based on average power of input signal over timespan, threshold to cut off noise
|
|
x = np.concatenate(history*history_diminish)*window/ATTENUATION
|
|
|
|
# estimate power spectral desity using autocorelate approach
|
|
psd = np.abs(np.fft.fft(np.correlate(x,x,'same')))[...,np.newaxis]
|
|
# frequencies
|
|
freqs = np.fft.fftfreq(psd.shape[0],1./RATE)[...,np.newaxis]
|
|
# frequency band vector _orthogonal_ to freqs
|
|
bands = np.logspace(np.log2(MIN_FREQ),np.log2(MAX_FREQ),EPIXELS+1,True,2)[np.newaxis,...]
|
|
# integrate energy within bands ~~ oh, oohhh: look at the orthoginality trick
|
|
bands = (freqs>bands[...,:-1]) & (freqs<=bands[...,1:])
|
|
energy = np.round(( bands * psd ).sum(0).squeeze())
|
|
#energy = np.round(( bands * psd / bands.sum(0) ).sum(0).squeeze())
|
|
|
|
# write some debug colorfull, very usefull
|
|
ansicolors = ('\033[30;1m%5.0f\033[0m', '\033[33;1m%5.0f\033[0m', '\033[1;31m%5.0f\033[0m')
|
|
sys.stderr.write('\r[%s] '%','.join(
|
|
(ansicolors[2] if k>400 else ansicolors[1] if k>200 else ansicolors[0]) %k for k in energy
|
|
))
|
|
|
|
for pixel in xrange(EPIXELS):
|
|
c = get_color(energy[pixel]) # consider using energy**0.5 instead
|
|
if SYMMETRIC == 1:
|
|
l.set_pixel(PIXELS / 2 + pixel, c[0], c[1], c[2])
|
|
l.set_pixel(PIXELS / 2 - (pixel + 1), c[0], c[1], c[2])
|
|
else:
|
|
l.set_pixel(pixel, c[0], c[1], c[2])
|
|
work = l.update()
|
|
# time.sleep(0.05)
|
|
|
|
if __name__ == '__main__':
|
|
try:
|
|
with_stream(loop)
|
|
except KeyboardInterrupt:
|
|
pass
|
|
|