fft

fft.py

#!/usr/bin/env python
import wave
import struct
import sys
import numpy as np
from math import sqrt
import matplotlib
matplotlib.use('Agg')
from matplotlib import pylab
import matplotlib.pyplot as plt

filename = sys.argv[1]

if __name__ == '__main__':
	# Open the wave file and get info
	wave_file = wave.open(filename, 'r')
	data_size = wave_file.getnframes()
	sample_rate = wave_file.getframerate()
	# print "Sample rate: %s" % sample_rate
	sample_width = wave_file.getsampwidth()
	duration = data_size / float(sample_rate)

	# Read in sample data
	sound_data = wave_file.readframes(data_size)

	# Close the file, as we don't need it any more
	wave_file.close()

	# Unpack the binary data into an array
	unpack_fmt = '%dh' % (data_size)
	sound_data = struct.unpack(unpack_fmt, sound_data)

	# Process many samples
	fouriers_per_second = 5 # Frames per second
	fourier_spread = 1.0/fouriers_per_second
	fourier_width = fourier_spread
	fourier_width_index = fourier_width * float(sample_rate)

	if len(sys.argv) < 3:
		length_to_process = int(duration)-1
	else:
		length_to_process = float(sys.argv[2])

	# print "Fourier width: %s" % str(fourier_width)

	total_transforms = int(round(length_to_process * fouriers_per_second))
	fourier_spacing = round(fourier_spread * float(sample_rate))

	# print "Duration: %s" % duration
	# print "For Fourier width of "+str(fourier_width)+" need "+str(fourier_width_index)+" samples each FFT"
	# print "Doing "+str(fouriers_per_second)+" Fouriers per second"
	# print "Total " + str(total_transforms * fourier_spread)
	# print "Spacing: "+str(fourier_spacing)
	# print "Total transforms "+str(total_transforms)

	lastpoint=int(round(length_to_process*float(sample_rate)+fourier_width_index))-1

	sample_size = fourier_width_index
	freq = sample_rate / sample_size * np.arange(sample_size)

	x_axis = range(0, 12)

	def getBandWidth():
		return (2.0/sample_size) * (sample_rate / 2.0)

	def freqToIndex(f):
		# If f (frequency is lower than the bandwidth of spectrum[0]
		if f < getBandWidth()/2:
			return 0
		if f > (sample_rate / 2) - (getBandWidth() / 2):
			return sample_size -1
		fraction = float(f) / float(sample_rate)
		index = round(sample_size * fraction)
		return index

	fft_averages = []

	def average_fft_bands(fft_array):
		num_bands = 12 # The number of frequency bands (12 = 1 octave)
		del fft_averages[:]
		for band in range(0, num_bands):
			avg = 0.0

			if band == 0:
				lowFreq = int(0)
			else:
				lowFreq = int(int(sample_rate / 2) / float(2 ** (num_bands - band)))
			hiFreq = int((sample_rate / 2) / float(2 ** ((num_bands-1) - band)))
			lowBound = int(freqToIndex(lowFreq))
			hiBound = int(freqToIndex(hiFreq))
			for j in range(lowBound, hiBound):
				avg += fft_array[j]

			avg /= (hiBound - lowBound + 1)
			fft_averages.append(avg)

	# MOUTH UPPER
	LOWER_MOUTH_MIN  = 768 * 4
	LOWER_MOUTH_MAX  = 2304 * 4
	UPPER_MOUTH_MIN  = 1008 * 4
	UPPER_MOUTH_MAX  = 2304 * 4

	result = []

	for offset in range(0, total_transforms):
		start = int(offset * sample_size)
		end = int((offset * sample_size) + sample_size -1)

		#print "Processing sample %i of %i (%d seconds)" % (offset + 1, total_transforms, end/float(sample_rate))
		sample_range = sound_data[start:end]
		## FFT the data
		fft_data = abs(np.fft.fft(sample_range))
		# Normalise the data a second time, to make numbers sensible
		fft_data *= ((2**.5)/sample_size)
		plt.ylim(0, 1000)
		average_fft_bands(fft_data)

		f_a = fft_averages
		f_a = [x**2 for x in f_a]
		mean = sqrt(sum(f_a))
		
		result.append(mean)
		# y_axis = fft_averages
		# """Stuff for bar graph"""
		# width = 0.35
		# p1 = plt.bar(x_axis, y_axis, width, color='r')
		# plt.title(mean)
		# """End bar graph stuff"""
		# filename = str('frame_%05d' % offset) + '.png'
		# plt.savefig(filename, dpi=100)
		# plt.close()
		
	MX = max(result)
	# print MX
	# print 'res'
	# print result
	for r in result:
		percent = int(r/MX*100)
		# print percent
		print '\tchannel_array[0] = %d; channel_array[1] = %d;' %  \
			(int(LOWER_MOUTH_MIN + percent*(LOWER_MOUTH_MAX - LOWER_MOUTH_MIN)/100),
			int(UPPER_MOUTH_MIN + percent*(UPPER_MOUTH_MAX - UPPER_MOUTH_MIN)/100))
		print '\tmaestro.setMultiTarget(2, 22, channel_array);'
		print '\tdelay(250);'
		
	# print "DONE!"

generated_code.c

  //generated_code from 9-Cat\ Excited_01.wav
  // 5 frame per second

	channel_array[0] = 4669; channel_array[1] = 5379;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3993; channel_array[1] = 4809;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 8048; channel_array[1] = 8231;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 4976; channel_array[1] = 5639;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3932; channel_array[1] = 4757;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3072; channel_array[1] = 4032;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 4239; channel_array[1] = 5016;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 6512; channel_array[1] = 6935;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3686; channel_array[1] = 4550;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3256; channel_array[1] = 4187;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3194; channel_array[1] = 4135;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3440; channel_array[1] = 4343;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 5959; channel_array[1] = 6468;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 6635; channel_array[1] = 7038;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3317; channel_array[1] = 4239;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 4300; channel_array[1] = 5068;
	maestro.setMultiTarget(2, 22, channel_array);
	//generated_code  from 9-Cat\ Excited_01.wav
	//delay 250ms and 
	delay(250);
	channel_array[0] = 4915; channel_array[1] = 5587;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3379; channel_array[1] = 4291;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 3379; channel_array[1] = 4291;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);
	channel_array[0] = 5898; channel_array[1] = 6416;
	maestro.setMultiTarget(2, 22, channel_array);
	delay(250);

run.sh

# need root privilige (sudo sh run.sh)
./arduino-serial/arduino-serial -b 9600 -p /dev/ttyACM0 -s s
aplay '9-Cat Excited_01.wav';

steps.md

Get and compile arduino-serial for send command via command line

git clone git@github.com:todbot/arduino-serial.git
cd arduino-serial
make
cd ..

Handle upload through arduino ide on ubuntu may cause error

avrdude: ser_open(): can't open device "/dev/ttyACM0": Permission denied
ioctl("TIOCMGET"): Inappropriate ioctl for device

solution:

sudo usermod -a -G dialout <replace_with_your_username>`
sudo chmod a+rw /dev/ttyACM0

Generate data

python fft.py 9-Cat\ Excited_01.wav > output

Paste into .ino file

run bash script to start

sudo sh run.sh