android-audio-recorder/docs/fft.py
Alexey Kuznetsov a4e3b13135 test fft
2016-04-02 16:20:02 +03:00

80 lines
No EOL
1.8 KiB
Python

import numpy as np
import matplotlib.pyplot as plt
import scipy.fftpack
import random
#
# https://web.archive.org/web/20120615002031/http://www.mathworks.com/support/tech-notes/1700/1702.html
#
def noise(y, amp):
return y + amp * np.random.sample(len(y))
# Fe = sample rate
# N = samples count
def plot(Fe, N, x, y):
plt.subplot(2, 1, 1)
print "power wav = %s" % np.sqrt(np.mean(y**2))
plt.plot(x, y)
plt.subplot(2, 1, 2)
yf = scipy.fftpack.fft(y)
NumUniquePts = np.ceil((N+1)/2)
# Bin 0 contains the value for the DC component that the signal is riding on.
fftx = yf[1:NumUniquePts]
mx = np.abs(fftx)
mx = mx / N
mx = mx**2
if N % 2 > 0:
mx[2:len(mx)] = mx[2:len(mx)]*2
else:
mx[2:len(mx)-1] = mx[2:len(mx)-1]*2
print "power fft = %s" % np.sqrt(np.sum(mx))
end = Fe/2
start = end / (N/2)
xf = np.linspace(start, end, N/2 - 1)
mx = np.sqrt(mx)
plt.plot(xf, mx)
plt.show()
def simple(Fe):
N = Fe
x = np.linspace(0.0, 1.0, N)
y = np.zeros(len(x))
y += 0.9 * np.sin(50.0 * 2.0*np.pi*x) + 0.5*np.sin(80.0 * 2.0*np.pi*x)
y += 0.6 * np.sin(20.0 * 2.0*np.pi*x) + 0.5*np.sin(80.0 * 2.0*np.pi*x)
y += 0.2 * np.sin(80.0 * 2.0*np.pi*x) + 0.5*np.sin(80.0 * 2.0*np.pi*x)
#y = noise(y, 2)
plot(Fe, N, x, y)
def real_sound_weave(durationMs):
Fe = 16000
N = Fe * durationMs / 1000
x = np.linspace(0.0, N, N)
y = np.zeros(len(x))
y += np.sin(2.0 * np.pi * x / (Fe / float(4500)))
y += 0.5 * np.sin(2.0 * np.pi * x / (Fe / float(4000)))
y += 0.5 * np.sin(2.0 * np.pi * x / (Fe / float(1000)))
y += 0.9 * np.sin(2.0 * np.pi * x / (Fe / float(7500)))
y += 1 * np.sin(2.0 * np.pi * x / (Fe / float(3000)))
m = np.max(np.abs(y))
y = y * 0x7fff
# y = y / m
#y = noise(y, 0x7fff)
plot(Fe, N, x, y)
#simple(1000)
real_sound_weave(100)